Clemson University
TigerPrints
All Dissertations
Dissertations
5-2015
Perceiving Soft Tissue Break Points in the Presence
of Friction
Bliss Altenhoff
Clemson University
Follow this and additional works at: http://tigerprints.clemson.edu/all_dissertations
Recommended Citation
Altenhoff, Bliss, "Perceiving Soft Tissue Break Points in the Presence of Friction" (2015). All Dissertations. Paper 1472.
This Dissertation is brought to you for free and open access by the Dissertations at TigerPrints. It has been accepted for inclusion in All Dissertations by
an authorized administrator of TigerPrints. For more information, please contact [email protected].
PERCEIVING SOFT TISSUE BREAK POINTS IN THE PRESENCE OF FRICTION
A Dissertation
Presented to
the Graduate School of
Clemson University
In Partial Fulfillment
of the Requirements for the Degree
Doctor of Philosophy
Human Factors Psychology
by
Bliss Altenhoff
May 2015
Accepted by:
Dr. Christopher Pagano, Committee Chair
Dr. Timothy Burg
Dr. Benjamin Stephens
Dr. Richard Pak
ABSTRACT
In minimally invasive surgery (MIS), surgeons face several perceptual challenges
due to the remote interaction with the environment, such as distorted haptic feedback
through the instruments due to friction produced from the rubber trocar sealing
mechanisms at the incision site. As a result, surgeons sometimes unintentionally damage
healthy tissues during MIS due to excessive force. Research has demonstrated that useful
information is available in the haptic array regarding soft tissues, which allows novices to
successfully perceive the penetration distance remaining until a material will fail based
on displacement and reactionary forces of simulated tissues using a haptic invariant,
Distance-to-Break (DTB). Attunement and calibration training was used in the current
study to investigate whether observers are able to identify material break points in
nonlinear compliant materials through haptic force application, while ignoring haptic
stimulation not lawfully related to the properties specifying DTB, including friction. A
pretest, feedback, posttest, and transfer-of-training phase design allowed participants to
probe four virtually simulated materials at varying levels of friction: no friction, low
friction, and high friction in the first experiment, and pull the simulated tissues in the
second experiment to investigate if perception of DTB generalizes to other tasks used in
MIS. Experiment 1 revealed that sensitivity to DTB can be improved through training,
even in the presence of friction, and that friction may assist observers to perceive fragile
tissues that otherwise would be below perceptual threshold. Experiment 2 revealed that
attunement and calibration to DTB also transfers to pulling motions.
ii
ACKNOWLEDGMENTS
I would like to thank Dr. Chris Pagano for being a wonderful advisor throughout
this process. He provided an abundance of excellent advice and guidance, while still
allowing me to pursue my own areas of interest in this project and work at my own pace.
I feel very fortunate to have been his student and don’t think there could have been a better place for me to spend my graduate years.
I would also like to thank my fellow members of the Perception and Action Lab,
especially Irfan Kil, for all the hours of engineering needed for this project and taking my
frantic phone calls when I was in need of technical help, and Leah Hartman for all the
advice and brainstorming on this journey with me.
I am also very grateful for the statistical advice and direction from Drs. Patrick
Rosopa and Dewayne Moore and Kristen Jennings, without whom I would not have been
able to successfully perform my first multilevel model analysis.
My committee members, Drs. Tim Burg, Ben Stephens, and Rich Pak have also
provided me with invaluable advice and recommendations, which helped steer the
direction of this project.
And of course, I am so thankful to my husband, Conner Altenhoff, for supporting
my decision to pursue my PhD, tolerating the late nights and weekends spent working,
and happily accompanying me to the South. Before moving here, I don’t think either of us expected to be so sad to see our time here come to an end.
iii
TABLE OF CONTENTS
Page
TITLE PAGE .................................................................................................................... i
ABSTRACT ..................................................................................................................... ii
DEDICATION ................................................................................................................iii
ACKNOWLEDGMENTS .............................................................................................. iv
LIST OF TABLES .......................................................................................................... vi
LIST OF FIGURES ....................................................................................................... vii
CHAPTER
I.
PERCEIVING SOFT TISSUE BREAK POINTS IN
THE PRESENCE OF FRICTION ........................................................... 1
MIS training and perceptual problems ..................................................... 2
Haptics: Tactile and kinesthetic info........................................................ 3
Direct perception and the haptic array ..................................................... 6
DTB.......................................................................................................... 8
Training perception of DTB................................................................... 10
Trocar friction ........................................................................................ 12
Purpose and goals .................................................................................. 14
II.
EXPERIMENT ONE ................................................................................... 17
Methods.................................................................................................. 18
Results .................................................................................................... 28
III.
EXPERIMENT TWO .................................................................................. 64
Methods.................................................................................................. 64
Results .................................................................................................... 65
IV.
GENERAL DISCUSSION .......................................................................... 82
iv
Table of Contents (Continued)
Page
Hypothesis 1........................................................................................... 82
Hypothesis 2........................................................................................... 83
Hypothesis 3........................................................................................... 84
Hypothesis 4........................................................................................... 86
Hypothesis 5........................................................................................... 87
Conclusions ............................................................................................ 88
REFERENCES .............................................................................................................. 91
APPENDICES ............................................................................................................... 97
A:
B:
Demographics Questionnaire ....................................................................... 98
Effects of Friction ........................................................................................ 99
v
LIST OF TABLES
Table
Page
1
Distance and reactionary force qualities at material
break point defining each simulated profile........................................... 28
2a
Average break point distance estimate means and
standard deviations (mm) by profile type in
experimental condition with no friction ................................................. 29
2b
Average break point distance estimate means and
standard deviations (mm) by profile type in
experimental condition with low friction ............................................... 29
2c
Average break point distance estimate means and
standard deviations (mm) by profile type in
experimental condition with high friction.............................................. 30
3a
Average break point distance estimate means and
standard deviations (mm) by profile type in
control condition with no friction .......................................................... 32
3b
Average break point distance estimate means and
standard deviations (mm) by profile type in
control condition with low friction ........................................................ 32
3c
Average break point distance estimate means and
standard deviations (mm) by profile type in
control condition with high friction ....................................................... 32
4a
Average R2, slope, and intercepts of simple regressions
predicting break point estimates from actual break
point during the pretest phase for each participant in
the experimental condition ..................................................................... 36
4b
Average R2, slope, and intercepts of simple regressions
predicting break point estimates from actual break
point during the feedback phase for each participant
in the experimental condition................................................................. 37
vi
List of Tables (Continued)
Table
Page
4c
Average R2, slope, and intercepts of simple regressions
predicting break point estimates from actual break
point during the posttest phase for each participant
in the experimental condition................................................................. 38
4d
Average R2, slope, and intercepts of simple regressions
predicting break point estimates from actual break
point during the ToT phase for each participant in
the experimental condition ..................................................................... 39
5a
Average R2, slope, and intercepts of simple regressions
predicting break point estimates from actual break
point during the pretest phase for each participant in
the control condition .............................................................................. 40
5b
Average R2, slope, and intercepts of simple regressions
predicting break point estimates from actual break
point during the feedback phase for each participant
in the control condition .......................................................................... 41
5c
Average R2, slope, and intercepts of simple regressions
predicting break point estimates from actual break
point during the posttest phase for each participant
in the control condition .......................................................................... 42
5d
Average R2, slope, and intercepts of simple regressions
predicting break point estimates from actual break
point during the ToT phase for each participant in
the control condition .............................................................................. 43
6
Estimates of foxed effects and standard error (SE) in
the experimental condition ..................................................................... 53
7
Estimates of foxed effects and standard error (SE) in
the control condition .............................................................................. 58
8
Break frequency occurrence in the ToT phase across
material and friction level for participants in the
experimental condition........................................................................... 63
vii
List of Tables (Continued)
Table
Page
9
Break frequency occurrence in the ToT phase across
material and friction level for participants in the
control condition .................................................................................... 63
10a
Average break point distance estimate means and
standard deviations (mm) by profile type in
control condition with no friction .......................................................... 65
10b
Average break point distance estimate means and
standard deviations (mm) by profile type in
control condition with low friction ........................................................ 66
10c
Average break point distance estimate means and
standard deviations (mm) by profile type in
control condition with high friction ....................................................... 66
11a
Average R2, slope, and intercepts of simple regressions
predicting break point estimates from actual break
point during the pretest phase for each participant ................................ 70
11b
Average R2, slope, and intercepts of simple regressions
predicting break point estimates from actual break
point during the feedback phase for each participant ............................ 71
11c
Average R2, slope, and intercepts of simple regressions
predicting break point estimates from actual break
point during the posttest phase for each participant............................... 72
11d
Average R2, slope, and intercepts of simple regressions
predicting break point estimates from actual break
point during the ToT phase for each participant .................................... 73
12
Estimates of foxed effects and standard error (SE) in
the experimental condition ..................................................................... 77
13
Break frequency occurrence in the ToT phase across
material and friction level ...................................................................... 81
viii
LIST OF FIGURES
Figure
Page
1
Relationship between an object’s distance and the size of its projection on the retina ................................................................... 8
2
Relationship between soft material displacement and
mechanical force required for that displacement ................................... 10
3
Schematic and photographic representation of the Core
Haptic Skills Training Simulator (Singapogu, et al.,
2013) ...................................................................................................... 20
4
Visual graphic used in calibration feedback phase
(Long et al., 2014). ................................................................................. 21
5
The four simulated material profiles and their
designated breaking point location ........................................................ 22
6
The four simulated material profiles and their
hypothetical breaking point location displayed
during the pretest and posttest phases .................................................... 24
7
The three simulated material profiles and their
designated breaking point location displayed during
the feedback training .............................................................................. 25
8
The four simulated material profiles and their
respective actual break point locations used in the
transfer-of-training phase ....................................................................... 27
9
Average pretest break point estimates as a function of
actual break point for all participants in the
experimental condition........................................................................... 30
10
Average posttest break point estimates as a function of
actual break point for all participants in the
experimental condition........................................................................... 31
11
Average pretest break point estimates as a function of
actual break point for all participants in the
control condition .................................................................................... 33
ix
List of Figures (Continued)
Figure
Page
12
Average posttest break point estimates as a function of
actual break point for all participants in the
control condition .................................................................................... 34
13
Pretest break point estimates as a function of actual
break point for all participants in the experimental
condition ................................................................................................ 54
14
Posttest break point estimates as a function of actual
break point for all participants in the experimental
condition ................................................................................................ 55
15
Pretest break point estimates as a function of actual
break point for all participants in the control
condition ................................................................................................ 59
16
Posttest break point estimates as a function of actual
break point for all participants in the control
condition ................................................................................................ 60
17
Average pretest break point estimates as a function of
actual break point for all participants..................................................... 67
18
Average posttest break point estimates as a function of
actual break point for all participants..................................................... 68
19
Pretest break point estimates as a function of actual
break point for all participants ............................................................... 78
20
Posttest break point estimates as a function of actual
break point for all participants ............................................................... 79
x
CHAPTER ONE
PERCEIVING SOFT TISSUE BREAK POINTS IN THE PRESENCE OF FRICTION
Minimally invasive surgical techniques offer patients the promise of smaller
incisions, reduced damage to the body, less pain, and shorter recovery times by inserting
long instruments and an endoscopic camera into small incisions via trocars (Breedveld &
Wentink, 2001; Bathea et al., 2004). Although traditional open surgeries allow the
surgeon to directly manipulate internal body organs and tissues through a large opening,
new technologies have allowed for a rise in minimally invasive surgery (MIS)
procedures, which require the surgeon to manipulate tissues indirectly through
laparoscopic tools and view the operation indirectly on a two-dimensional monitor. Due
to the challenging visual conditions and the unintuitive nature of tool manipulation with
MIS, a large amount of training is required to develop these skills, which is often
performed with the use of simulators. In fact, there are more surgical training simulators
available for laparoscopic training than any other type of medical training task (Coles,
Meglan, & John, 2011).
Research suggests that surgeons sometimes unintentionally damage healthy
tissues during MIS procedures. For example, in 60 simulated cholecystectomies
performed by 60 surgical trainees, use of excessive force was the third most common
type of error committed, with 187 occurrences recorded during the 60 procedures (Tang,
Hanna, & Cuschieri, 2005). All instances of tissue damage were determined to be the
result of excessive force. Injuries of bile ducts during a cholecystectomy occur three
times more often in laparoscopic surgery than in open surgery (Archer, Brown, Smith,
1
Branum, & Hunter, 2001; Traverso, 1999). Out of approximately 500,000 patients
receiving a laparoscopic cholecystectomy each year, about 1,500 to 2,000 will experience
damage to the bile ducts (Hugh, 2002). A great majority of these damages during
cholecystectomy are caused by errors in perception, rather than errors in skill,
knowledge, or judgment, with surgeons injuring unseen bile ducts during dissection or
deliberately cutting a bile duct when he or she believed it to be something else (Way et
al., 2003).
MIS training and perceptual problems
Despite documented benefits of using MIS methodologies rather than open
surgery, surgeons face several perceptual challenges with MIS similar to other
environments where an operator is controlling or viewing an object remotely. For
example, humans struggle to perceive depth and size of objects in virtual or remote
environments, underestimating egocentric distances (0m – 30m) by as much as 50%
(Altenhoff et al., 2012; Napieralski et al., 2011; Richardson & Waller, 2005; Thompson
et al., 2004; Witmer & Kline, 1998). Tittle, Roesler, and Woods (2002) have termed
these difficulties “the remote perception problem.” Unlike open surgery, where a surgeon can look directly into the operative scene and see his/her hands and tools
manipulating the scene, during MIS the surgeon views a monitor, which produces a
mislocation of the endoscope and prevents the surgeon from being able to simultaneously
observe his/her hands and the scene. This difference between the endoscope’s viewpoint and the surgeon’s viewpoint as if he/she were looking directly into the abdomen produces
a variety of perceptual challenges that the surgeon must overcome. Because the
2
instruments rotate around the incision point through a trocar, surgeons must translate a
mirrored image produced during sweeping, side-to-side motions, which is also magnified
on the monitor, further contributing to the scaling challenges as experienced in other
remote environments. Depending on how far the instrument is inserted, the surgeon’s movements may produce a magnified or reduced effect on the instrument tip.
Additionally, the endoscope is usually controlled by someone other than the surgeon,
which can cause the endoscope’s viewpoint to be different from what the surgeon’s would be if he/she looked down into the abdomen (Breedveld & Wentink, 2001). In
addition to viewing a rotated view of the surgery, surgeons are passively viewing the
scene as the assistant controls the camera, which breaks the perception-action link present
in regular environments (Tittle et al., 2002; Gomer, Dash, Moore, & Pagano, 2009).
Haptics: Tactile and Kinesthetic Info
Not only does the remote perception problem in MIS procedures make it difficult
for surgeons to accurately perceive visual information on the monitor, but the
arrangement also creates distorted haptic feedback through the surgeon’s instruments (Den Boer et al., 1999; Van den Dobbelsteen, Schooleman, & Dankelman, 2007). Just as
the surgeon cannot directly see the tissues he or she is operating on, organs and tissues
are touched indirectly and softness is assessed with surgical tools rather than his or her
hands. The haptic feedback a surgeon receives in open surgery when directly touching
soft tissue consists of both tactile information (cutaneous stimulation) and kinesthetic
information. When the tissue or organ is touched or squeezed, the sense of touch, or
tactile information, is perceived via sensory receptors in the finger tips. Kinesthetic
3
information is the sense of one’s body position and movement, wielded objects, and
probed surfaces communicated by receptors in joints, tendons, muscles, and skin (Loomis
& Lederman, 1986, Pagano & Cabe, 2003; Pagano, Carello & Turvey, 1996; Pagano &
Donahue, 1999; Perreault & Cao, 2006; Srinivasan & LaMotte, 1995; Turvey, 1996).
Together, tactile and kinesthetic information make up a person’s haptic sense. Though both haptic and kinesthetic information both play a role in the surgical task, kinesthesis is
primarily responsible for providing information about the interactions between the distal
ends of the tools and the properties of the tissue.
As surgeons apply force onto body organs and tissues, they can immediately
obtain useful haptic information based on the compliance of the tissue. However,
because tactile information has been shown to be useful when discriminating between
deformable materials (Srinivasan & LaMotte, 1995), surgeons must alter how they judge
compliancy of materials with only indirect contact through the tools. Relying primarily
on kinesthesis, the amount of tissue displacement or resistance in response to the amount
of force applied reveals useful property information about the compliancy, or softness, of
a tissue to the surgeon (Bergman Tiest & Kappers, 2009; Vincentini & Botturi, 2009;
Srinivasan & LaMotte, 1995). This compliancy and stiffness information gained from
force feedback gives information about fragility and may specify motor adjustments
necessary to avoid damaging materials. Although vision can offer some clues about
compliancy as tissue is displaced in response to forces, we know that the visual
information available in MIS is also distorted by unnatural camera angles, ambiguous
scaling, and lack of stereoscopic information. Even if the operation were viewed directly,
4
vision does not provide any information about force being applied or reactionary force
from the tissue, which is necessary to determine the compliancy of a material. Vision
alone does not provide enough information to accurately judge softness or fragility of a
material (Srinivasan & LaMotte, 1995; Klazky, Lederman, & Matula, 1993; Smyth &
Waller, 1998).
It is important that surgeons are trained to pick on useful haptic information in
this new environment. Evidence suggests that haptic feedback does improve performance
in both MIS tasks and basic laparoscopic training tasks, particularly when pushing or
pulling (Chmarra, Dankelman, van den Dobbelsteen, & Jansen, 2008). When performing
a dissection task robotically without force feedback, gynecologic residents participants
applied 50% more force than with force feedback and committed three times the number
of injury-causing errors (Wagner, Stylopoulos, & Howe, 2002). Particularly, research
demonstrates that receiving haptic feedback in a virtual training environment may be
especially critical during early training phases for psychomotor skill acquisition (Ström et
al., 2006). Due to the indirect contact with the tissue, future surgeons have to learn about
force feedback before they can safely conduct actual surgery. For example, during
surgery the operator may perceive forces 0.2-4.5 times the force generated. Realistic
simulators with haptic feedback are thought to lead to better overall performance, faster
learning, and high transfer of skill to operating on actual tissue (van der Meijden, &
Schijven, 2009). Some warn that learning tasks in VR without realistic haptic feedback
may result in negative learning effects when these tasks are completed on actual tissue,
where appropriate application of force plays an important role in surgical performance
5
(Chmarra et al., 2008; Srinivasan & LaMotte, 1995). Despite extensive training,
typically developing expert eye-hand coordination skills, no haptic skills training in
detecting tissue break point or force perception is required.
Direct Perception and the Haptic Array
Similar to the well-studied lawful relationships available in the optic array of
looming and time-to-contact (Lee, 1976; Hecht & Savelsbergh, 2004), research has also
demonstrated that information is available in the haptic array of soft tissues that specifies
to a surgeon when a tissue will break (Long et al., 2014). As surgeons apply a given
amount of force to a soft tissue, the resulting displacement grows less and less as they
probe deeper, indicating that the tissue is becomingly increasingly stiff and that it may
soon break. This changing compliancy in human tissue is often the result of a lawful,
biomechanical relationship (Brouwer et al., 2001; Carter, Frank, Davies, McLean, &
Cuscheri, 2001; Fung, 1993; Rosen, Brown, De, Sinanan, & Hannaford, 2008; Yamada,
1970) – one which surgeons may be able to attune to in order to perceive how much
farther they can probe before breaking the tissue. If trained to accurately interpret these
lawful relationships, surgeons should better understand the structural capacity of human
tissues, allowing them to apply more appropriate forces and reduce trauma or breakage of
healthy tissues.
Based on research and theories of J. Gibson (1950, 1966, 1979), many have
studied ways in which the human perceptual system is able to attune to information
available in an ambient stimulus array (Long, et al., 2014; Cabe & Pittenger, 1992; Cabe,
2011), particularly in the optic array (Gibson 1966, 1979; Lee, 1976; Bingham & Pagano,
6
1998). According to Gibson (1966, 1979), information in the optic/haptic array that
follows the laws of physics reveals invariants to an observer, which can be used to guide
actions. For example, looming of an object in the environment on an observer’s retina can be perceived through information in the light. As the distance between the object and
the observer decreases, the rate of expansion specifies to the observer the time until one
will make contact with the object, known as time-to-contact (TTC). Lee (1976)
demonstrated that TTC provides actionable information to an observer, based on the
relationship between the distance between the object and the observer (area on the retina)
and velocity (rate of change on the retina), which he labeled tau (see Figure 1). For
example, as an observer approaches a stop sign, the sign subtends a larger and larger
amount of space in the visual field. As the area of the sign increases on the retina, the
observer perceives the distance between themselves and the sign to decrease. This
relative rate of expansion of the sign is expressed as:
Relative Rate of Expansion = ΔArea/ΔTime
Area
The inverse specifies TTC, which denotes time remaining until the distance between the
observer and the sign reaches zero, and is expressed as:
TTC = ___Area___
ΔArea/ΔTime
7
Figure 1. Relationship between an object’s distance and the size of its projection on the
retina.
If the observer is sensitive to TTC, they can perceive the time remaining before
the distance reaches zero without computing lower-order variables such as velocity,
object size, or object distance. Since, researchers have also examined looming and
similar relationships in other modalities, such as acoustic TTC (Shaw, McGowen &
Turvey, 1991), time-to-topple based on haptic information (Cabe & Pittenger, 1992),
impending contact based on acoustic information (Schiff & Oldak, 1990), and haptic
looming (Cabe, 2011). Thus, it is likely that relationships similar to TTC exist in the
haptic array when approaching the break point while deforming a soft tissue.
Distance-To-Break (DTB)
Tissue compliancy is perceived through the amount of tissue displacement or
resistance in response to the amount of force applied, which gives the surgeon
8
information about fragility (Bergman Tiest & Kappers, 2009; Vincentini & Botturi, 2009;
Srinivasan & LaMotte, 1995), and may offer information specifying the distance
remaining until the material breaks. Many soft biological tissues follow an exponential
stress-strain pattern (Brouwer et al., 2001; Carter, Frank, Davies, McLean, & Cuscheri,
2001; Fung, 1993; Rosen, Brown, De, Sinanan, & Hannaford, 2008; Yamada, 1970)
where the reactionary forces increase in an exponential fashion as the displacement into
the tissue increases towards the point of failure until the tissue finally breaks as the
structural limits are reached. Long et al. (2014) demonstrated that participants can
successfully perceive the penetration distance remaining until a material will fail based
on displacement and reactionary forces of simulated tissues using a haptic invariant
comparable to TTC, Distance-to-Break (DTB).
DTB = _____Force______
ΔForce/ΔDisplacement
Similar to TTC, DTB is a higher-order parameter that does not require the
computation of lower-order variables such as force or tissue stiffness. Rather, it is the
ratio between the amount of force applied and the change in reactionary force as
displacement increases. As force is applied, deformation of the material and reactionary
forces specify to the operator the amount of displacement which the tissue can tolerate
before breaking (See Figure 2). Using nine different tissue profiles with varying break
point distances and varying forces required to break each material, Long et al. (2014)
demonstrated that novice participants are sensitive to DTB and can improve the
perceptual skill of judging break points with a brief training phase through attunement
and calibration.
9
Figure 2. Relationship between soft material displacement and mechanical force required
for that displacement.
Training Perception of DTB
Virtual environments (VE’s) are a common means for providing training for situations that are dangerous, expensive, rare, or remote, such as laparoscopic surgery
training (Bliss, Tidwell, & Guest, 1997; Darby, 2000; Peters et al., 2008). A main
advantage of virtual environments is that they provide a controlled scenario so users can
repeatedly and safely interact with situations. Due to the nature of MIS tasks, training on
actual patients would be too dangerous and though cadavers and animal tissues are
sometimes an option, it is often expensive, allowing the training surgeon minimal errors
before being rendered useless (Coles, Meglan, & John, 2011). However, medical
simulators are becoming an increasingly accepted tool for the extensive training
necessary to prepare surgeons. Medical simulators provide a safe, yet realistic
10
environment in which the surgeon can practice a task repeatedly and frequently to help
maximize learning with the freedom to make mistakes.
With practice, it may be possible to increase sensitivity to DTB by improving
observers’ ability to discriminate the useful and meaningful properties available within
the haptic array (Gibson, 1953; 1963; 1969; Gibson & Gibson, 1955). Virtual simulators
provide an opportunity for novices to experience frequent, repeated haptic interaction
with tissues, providing an ideal scenario for perceptual learning (i.e. training). Sensory
systems are continually exposed to limitlessly rich information available for haptic
perception that may or may not convey useful perceptual information about object
properties. Through experience and feedback, perceivers attend to the useful
information, and haptic perception becomes ‘tuned’ to the mechanically useful properties lawfully related to perceptual variables, known as specifying variables (Wagman,
Shockley, Riley, & Turvey, 2001; Withagen & Michaels, 2005). Before feedback,
perceivers perceptually estimate an object property based on a combination of variables,
both specifying and non-specifying variables, which are ambiguously related to the
property. Referred to as “education of attention” or “attunement”, observers learn to converge on the variables that are most correlated with the object property and which
accurately predict it, attuning to the salient perceptual invariants that specify useful
information (Gibson & Gibson, 1955; Gibson, 1963; Withagen & Michaels, 2005).
The theory is that learning is more efficient when it involves perceptual
attunement to meaningful information as opposed to the acquisition of complex mental
structures. Over time, the perceptual system’s output is also correctly scaled for accurate
11
perceptual judgments, resulting in calibration of haptic perceptual systems (Withagen &
Michaels, 2005). Previous research in our lab has found such performance
improvements, with accuracy of participant force applications during probing, grasping,
and/or sweeping movements improving after experiencing a training phase, which
incorporates visual feedback (Singapogu et al., 2011, 2013, in press; Long et al., 2012,
2014). Attunement and calibration training will be used in the current study to improve
the ability of observers to perceive material failure points.
Trocar Friction
Although Long et al. (2014) were able to demonstrate that perceiving DTB is a
trainable perceptual skill, several other factors must be taken into consideration to make
sure the training transfers to actual MIS procedures before a DTB training program can
be fully developed. For example, basic research needs to demonstrate that training on the
Core Haptic Skills Simulator transfers to perceiving DTB in real human tissues. Also,
more research is needed to investigate how participants learn to perceive DTB with the
other motions performed during MIS surgery such as pulling, sweeping, and grasping
(Singapogu et al., 2012b) and with haptic sensations more representative of the forces
generated by multiple interactions within the MIS environment. Not only do surgeons
receive force feedback from applying pressure to internal organs and tissues, but the
trocars, abdominal wall, and mass of the instrument alter the haptic feedback to the
surgeon as well (Picod, Jambon, Vinatier, & Dubois, 2004). Trocars are a sealing
mechanism, made of short tubes at the site of the incision that act as a portal for tools to
access the body organs and/or tissues and often maintain pressure within the body cavity
12
when fluid or gas is pumped in to better expose the surgical site.
Because injury inflicted on tissues during MIS tends to be more frequent than in
open surgery, it is imperative that surgeons learn to reduce the inappropriate overapplication of forces which damage healthy tissues. It’s possible that many surgeons learn to ignore many of the complex haptic forces and compensate by visually observing
the deformation of tissue as force is applied. However, vision will not always provide
enough information to a surgeon to perceive when a tissue is about to fail (Srinivasan &
LaMotte, 1995; Klazky, Lederman, & Matula, 1993; Smyth & Waller, 1998), particularly
if an unhealthy or abnormal tissue visually appears to be normal. As discussed by Way et
al. (2003), a great majority of these damages during cholecystectomy are caused by errors
in perception, rather than errors in skill, knowledge, or judgment. Although the surgeon
is operating indirectly on a patient and challenged with issues of the remote perception
problem, some believe the friction generated by the rubber seal in the trocar to be the
most significant contributor to perceptual challenges which surgeons are sometimes not
able to overcome (Perreault & Cao, 2006).
In order to accurately attune to information of the biomechanical properties
inherent in DTB, surgeons must be able to ignore haptic stimulation that is not relevant to
the properties that are trying to perceive (i.e. non-specifying variables), including friction.
But because the friction of the trocar can be relatively large compared to from the forces
associated with the interactions between the tool and tissue, surgeons may not perceive
that a tissue is near failure if it is very compliant or fragile. One may have to probe
harder in order to perceive the compliancy of a soft tissue, especially if the haptic
13
feedback from friction is very high. Friction varies based on the type of trocar, the
movement velocity and direction of the surgeon’s pushing or pulling gestures, and
moistness (van den Dobbelsteen, Schooleman, & Dankelman, 2007).
Perreault and Cao (2006) demonstrated that trocar friction may cause an increased
haptic perception threshold, with novices applying more force and taking more time to
detect contact with tissue when friction is present. Early in training, surgeons must learn
how to operate despite this challenge. It’s possible that some surgeons are already attuning to DTB or other haptic skills to differentiate between useful haptic information
and non-specifying variables, while others choose to rely on visual information about
deformation. It is hypothesized that people can learn to ignore trocar friction, similar to
the way the human perceptual system has been shown to accurately perceive the length of
a rod by attuning to the invariant of inertia, ignoring the effects of wielding in different
media such as air and water (Pagano & Cabe, 2003; Pagano & Donahue, 1999). Also,
Lamata et al. (2008) demonstrated that surgeons were able to discriminate between four
different tissues with only force information available, despite large amounts of force
feedback from trocar friction. The goal of this research is to train participants how to
attune to DTB and ignore other non-specifying variables so they can appropriately make
use of perceptual information from visual and haptic modalities.
Purpose and Goals
Experienced surgeons have demonstrated the skills to accurately produce and
perceive haptic forces, although it is unlikely that they were specifically trained how to
attune to those forces. Training devices are currently being developed that are
14
specifically devoted to training haptic skills. Trainees have shown significant
improvement even after only a brief training period, demonstrating that it is a learnable
skill (Singapogu et al., 2012a; Singapogu et al., 2012b). Although few experiments have
investigated the effect of haptic feedback in a virtual environment (VE) simulator, the
majority of research supports the idea that haptic feedback should be incorporated into
VE training based on findings on the importance of haptics in minimally invasive surgery
(Singapogu et al., 2012a).
Two experiments are designed to investigate whether observers are able to
perceive DTB in nonlinear compliant materials through haptic force application, even
with a simulated friction term added to the force feedback, and then use this information
to identify the distance remaining until mechanical failure. The first experiment will test
four hypotheses:
1. Participants are sensitive to DTB
2. Participants can detect DTB with varying friction levels present
3. Ability to locate DTB with varying levels of friction is a skill that can be
improved through training.
4. Sensitivity to DTB will transfer to a task where participants must stop before
the break point is reached
The second experiment will test one additional hypothesis:
5. Sensitivity to DTB generalizes to a pulling task.
15
Procedure and materials used are very similar to those used by Long et al. (2014).
The first experiment examined to see if participants’ perception of DTB is altered by the
presence of varying levels of simulated trocar friction. Similar to Pagano and Cabe (2003;
Pagano & Donahue, 1999), which demonstrated that participants were able to attune to a
mechanical invariant, inertia, even when other forces (e.g. water resistance) were
included, we expect novices to attune to DTB of different nonlinear materials while
ignoring other forces. This experiment also examined participants’ ability to improve
perception of DTB with training. It is possible to increase an observer’s reliance on perceptual invariants with feedback and practice attuning to the relevant information in
the stimulus array (E. Gibson, 1969; J. Gibson, 1966; Withagen & Michaels, 2005). With
experience, the useful information will become more distinct within the haptic array as
the perceptual system identifies them as being lawfully related to material properties of
the tissue predicting failure. Once an observer becomes more sensitive, or attunes, to the
relevant mechanical properties, continued feedback will allow the haptic perceptual
system to calibrate, becoming more sensitive to the useful mechanical properties as
useful information is scaled for accurate perceptual judgments. Attunement and
calibration have been shown to improve perceptual judgments of kinesthetic properties
through training and feedback (Long, et al., 2012, 2014; Singapogu, et al., 2013, 2014;
Wagman et al., 2001; Withagen & Michaels, 2005). With training, attunement and
calibration should allow participants to become sensitive to the mechanical information
specifying the location of material failure points and improve judgments of DTB.
16
CHAPTER TWO
EXPERIMENT ONE
To test for effects of attunement and calibration, probing of simulated materials
was evaluated with a pretest, feedback, posttest, and transfer-of-training phase design,
with performance data in the pretest addressing Hypothesis 1, that participants are able to
detect DTB with friction present, and performance data in the posttest and transfer-oftraining phases addressing Hypothesis 2, that DTB is a trainable skill that participants can
calibrate with training. Friction was simulated in some trials during each of the four
phases. To allow free exploration of probing materials without revealing feedback of
break point judgments during pretest and posttest phases, materials were designed to
provide useful information as the participant rode the nonlinear curve of the
biomechanical properties of the tissue, but not actually break at the breaking point.
Rather, at the point of mechanical failure, the tissues’ force profile “flattened out” by
maintaining forces as they were presented at the point of failure. During the feedback
phase, feedback was provided visually. To investigate effects of DTB attunement in a
more realistic MIS scenario, further validating training capability of the Core Haptic
Skills Trainer, the transfer-of-training phase presented virtual materials to participants
that actually “broke” at the appropriate breaking point. These two types of presentations are referred to as Task 1 and Task 2.
Task 1 is an exploratory break detection phase which allows participants to freely
explore various simulated materials by pushing, or probing, into the material with a
laparoscopic tool. Similar to other training simulators, the Core Haptic Skills Simulator
17
allows participants to explore the theoretical materials with great variety of force
applications, breaking materials numerous times, without the negative consequences as in
actual surgery. Participants were encouraged to examine each tissue to indicate the
location at which they believe it feels as if it should break, applying forces both greater
than and less than what the virtual mechanical could withstand. Task 2 presented virtual
nonlinear materials that truly “broke”, examining if participants could successfully detect
DTB without breaking the simulated material. During the transfer-of-training phase,
Task 2 instructions were used, instructing participants to probe as close as possible to the
breaking point without actually breaking the material. In order to successfully complete
this task, participants had to perceive the location of the material’s breaking point before
actually tearing the material.
Methods
Participants
50 university undergraduate students between the ages of 17 and 22 (M = 18.2,
SD = 0.7) participated in Experiment 1 after providing informed consent, none of whom
had any experience practicing MIS. 35 were female and 15 were male. Participants
received course credit in exchange for their participation.
Materials & Apparatus
1. Simulator
Nonlinear soft tissues were rendered using the Core Haptic Skills Trainer, a
simulator developed at Clemson University with the purpose of training force-based
skills in laparoscopic surgery. The simulator emulates three different force-based skills
18
identified as particularly salient in minimally invasive surgery; grasping, probing, and
sweeping (see Singapogu et al., 2011, 2012a, 2012b, 2013). Probing was used in the
present study.
The force-based skills were integrated into a comprehensive simulator containing
a single input device permitting the user to make discrete probing, grasping, and
sweeping motions (see Figure 3). The input device was a laparoscopic surgical forceps
tool with a scissor grip handle with pinchers removed (a Covidien AutosutureTM Endo®
device, Dublin, Ireland). A robotic motion system delivered force feedback to the input
device through two direct-drive DC motors (Tohoku RiochTM, Miyagi 987-0511, Japan)
located at the center and the end of the forceps shaft. Through a series of computer
algorithms, the system renders force feedback by generating a torque in response to user
motion.
Haptic feedback rendered by the simulator emulates the tool coming into contact
with and encroaching into an amenable mass, such as soft tissue. For probing, the user
applies force through the input device by gripping the handles of the input device and
pushing the tool forward. Advancing the tool produced feedback imitating coming into
contact with and then pushing onto soft tissue, effectively simulating the tensile forces
experienced as one stretches soft tissue.
Task 2 is designed to present haptic feedback which would render the simulated
material truly ‘breaking,’ or failing, when excessive force is applied. As the user applies more force through the input tool, resistive force feedback increases at an exponential
19
rate. Once the applied force becomes great enough, resistive feedback will immediately
cease, emulating a soft tissue perforation.
Figure 3. Schematic and photographic representation of the Core Haptic Skills Training
Simulator (Singapogu, et al., 2013).
2. Visual feedback
Visual feedback was incorporated into the feedback training phase, allowing
participants to view errors and then adjust, or calibrate, their force application after each
trial. The feedback was presented by a custom visual graphic displayed on a monitor,
which indicates tool position and placement along the simulated material (see Figure 4).
The graphic included a movable red vertical, dynamic bar indicating normalized probed
distance and a fixed blue vertical bar indicating the actual break point position. The red
marker was proportional to the placement of the tool, and moved in response to
increasing and decreasing applied force through the surgical input tool. At the starting
position, the marker was located at the far left, moving from left to right as force is
applied. Because the breaking point for each simulated material varied, relative to the
material profile itself (described in detail below), the indication for break point in the
graphic was static. Thus, the location of the break point in the graphic did not change;
20
only the application force required to move the indicating marker varied. Participants
were asked to make their estimate of where they would place the tool to arrive as close to
the designated break point as possible without going past. Once a participant verbally
indicated they had made their estimate, they held the position of the tool while the
experimenter made the graphic display available so the participant was able to view
where their estimate was located relative to the break point. Participants were then asked
to adjust the tool as necessary to align it with the break point.
Figure 4. Visual graphic used in calibration feedback phase (Long et al., 2014).
3. Simulated Material Profiles
Four different nonlinear materials were simulated, based on profiles similar to soft
tissues exhibiting exponential stress-strain relationships in response to compressive and
tensile force loadings (Brouwer et al., 2001; Fung, 1993; Rosen, et al., 2008). The four
compliance profiles and breaking points were designed to be the product of two different
material strengths (F) at four different displacement locations (d) (see Figure 5). Thus,
each material contained a different point of failure, or location at which it would ‘break.’ Observers could not rely solely upon one varying dimension or the other when correctly
determining DTB, but must rely on the invariant relationship between the two of them.
21
As one dimension is modified and the point of failure changes, the relationship will still
be maintained, which should be sufficient for specifying DTB.
Figure 5. The four simulated material profiles and their designated breaking point
location.
Each of the four materials were presented with varying amounts of simulated
trocar friction: no friction (0N), low friction (1.5N), and high friction (3N). This range of
friction levels encompasses the actual trocar friction observed, which can range from 0.25
N to 3 N (van den Dobbelsteen, Schooleman, & Dankelman, 2007). Thus, during pretest,
posttest, and transfer-of-training phases, participants examined each material with no
added friction, with low levels of trocar friction, and high levels of trocar friction.
During the calibration phase, only three materials were used (see Figure 7). Tables 1a, 1b,
and 1c display all of the metrics defining the nonlinear characteristics for each material
profile, including break point distance and reactionary force.
Procedure
22
This experiment used a pretest, feedback, posttest, transfer-of-training model to
examine attuning and calibration effects to DTB. The pretest provided a pre-training
baseline to compare with posttest performance after feedback training. An additional
transfer task evaluated the degree to which DTB perceptual skill carried over to a novel,
more realistic MIS task. On the initial day of testing, participants who provided informed
consent completed the pretest after an introductory training phase, which allowed
participants to survey a single nonlinear material, helping them become comfortable with
the laparoscopic tool and the task. Within the next seven days, participants returned for
the feedback training phase, posttest, and transfer-of-training phases.
1. Pretest Phase
For the first phase, participants completed what has been defined as Task 1. They
applied forces up to and beyond a hypothetical break point for four simulated materials
presented at 3 varying levels of simulated trocar friction (no friction, low friction, and
high friction), with the goal of identifying the location at which the material should fail
(see Figure 7). Each material was presented three times with no friction, three times with
low added friction, and three times with high added friction (4 materials x 3 friction
levels x 3 presentations), for a total of 36 trials. No visual feedback was provided and
once the break point was reached, the material did not actually break so as to minimize
haptic feedback indicating successful judgment of DTB. They could freely explore the
material at whatever speed they feel most comfortable. Participants made their estimates
by suspending their force application and verbally designating their estimate to the
experimenter. The experimenter logged the trial in MatLab, which captured the applied
23
force and distance at that moment. At the end of each trial, the participant returned the
surgical tool to the starting position before beginning the next trial.
Figure 6. The four simulated material profiles and their hypothetical breaking point
location displayed during the pretest and posttest phases.
2. Feedback training phase
When participants returned for the second day of the experiment they completed
the training phase, which followed the same procedures as the pretest, but incorporated
the visual feedback graphic to allow participants to calibrate their haptic estimate, and
utilized for only three of the four experimental tissue profiles at two levels of friction (3
materials x 2 friction levels x 4 presentations) for a total of 24 trials. The feedback
training phase was completed two to eight days after the pretest phase (M = 4.1, SD =
1.8). Participants were informed that the goal of the training is to learn to apply sufficient
force onto each simulated profile without ‘breaking’ the material. Similar to the pretest, participants were allowed to freely explore the material at any speed or direction. They
were also instructed that identifying the failure point should occur before reaching the
breaking point, and that later phases will score excessive force applications as an error.
24
Similar to the pretest, participants indicated the location of the hypothetical
breaking point, although rather than immediately returning to the starting position for the
next trial, they were then shown a visual graphic of their performance, allowing them to
calibrate and make adjustments to their haptic estimate (see Figure 4). The task was to
locate the designated breaking point along the three materials depicted in Figure 7, again
applying the amount of force they believed was required to puncture, or break, the
material. After receiving feedback, the participant was allowed to adjust the tool to feel
the appropriate ‘break’ point. In order to examine practice effects, half of the participants
were randomly assigned to participate in a control condition, in which they completed the
same task as the pretest during what would be the feedback phase for those in the
experimental condition, resulting in a pretest, pretest, posttest, transfer-of-training model.
Figure 7. The three simulated material profiles and their designated breaking point
location displayed during the feedback training.
3. Posttest phase
Participants took a five-minute break between concluding the feedback training
phase and beginning the posttest phase, which used the same protocol and the same four
materials as the pretest phase. As in the pretest, each material was presented three times
25
with no friction, three times with low added friction, and three times with high added
friction (4 materials x 3 friction levels x 3 presentations), all without any visual feedback,
for a total of 36 trials.
4. Transfer-of-Training (ToT) Task
Participants took another five-minute break between concluding the posttest phase
and beginning the transfer-of-training phase. The transfer task was similar to the first
three phases, presenting the same four materials as in the pretest and posttest, except that
the designated break point location within the simulated profiles was rendered to truly
emulate breakage. As force was applied, the reactionary force of the material increased
until a certain point at which the material failed (see Figure 8), haptically emulating
puncture. The same four tissue profiles used in the pretest and posttest phases were used
for this task, with the breaking points occurring at the same displacement location,
although the material function approaches an asymptotic direction at the break point.
Each participant was instructed to apply as much force as they could to the materials
without breaking the material. During the instructions, they were given an analogy to
better understand the task: like being near the edge of a cliff, the goal was to inch as close
to the edge as possible without going over. Any trials in which a material was broken
with excessive force were marked as an error, ending the trial, which were then
represented at the end of the 36 original material presentations. Participants repeated
trials of broken materials until they successfully complete the 36 trials (4 materials x 3
friction levels x 3 presentations). Performance was measured based on the proximity of
force application to the breaking point and the number of tissue breaks.
26
Figure 8. The four simulated material profiles and their respective actual break point
locations used in the transfer-of-training phase.
Metrics for Analysis
1. Distance
Displacement traveled by the input device into the simulated materials was
presented by the simulator in terms of millimeters, which ranged from 0 to 35 mm, with
four values designed as breaking points at 7.5, 15, 22.5, and 30 (see Table 1).
2. Force
Reactionary force rendered by the simulator was presented as rendered voltage
and transformed into Newtons, both of which are displayed in Table 1. Two voltages
will define the reactionary behavior by the simulator: 3 and 5. The simulator will directly
record voltage, which is then transformed into Newtons (see Table 1).
27
Table 1
Distance and reactionary force qualities at material break point defining each simulated
profile
Material
Profile
1
2
3
4
Distance – all
friction levels
Millimeters
7.5
15
22.5
30
Reactionary force
– no friction
Newtons
3
5
3
5
Reactionary force
– low friction
Newtons
4.5
6.5
4.5
6.5
Reactionary force
- high friction
Newtons
6
8
6
8
3. Accuracy
Accuracy will be defined as the difference between the participants’ indicated
break point location and the actual break point location of the simulated material profiles.
When presented with materials that do not truly break, as in Task 1, the difference could
be positive, indicating over application of force, or negative, indicating under application
of force. When presented with materials that truly break with excessive force
application, as in Task 2, accuracy will only be negative since estimates must be short of
the true break location.
Results
Data were screened for outliers and for logging errors with the simulator. Due to
the restricted range of motion of the simulator, no trials exceeded a z-value of +3, so no
trials were excluded as outliers. However, 20 pretest trials and 27 posttest trials in the
experimental condition, as well as 20 pretest trials and 20 posttest trials in the control
condition, were not correctly recorded, with all values logged as 0 and were discarded.
28
Performance was assessed by analyzing displacement into the simulated material
via distance in millimeters. Means and standard deviations of distance are displayed by
material type in the experimental condition in Tables 2a, 2b, and 2c. Break point
estimates from the pretest and posttest, averaged across all participants in the
experimental group are also displayed in Figures 9 and 10.
Table 2a
Average break point distance estimate means and standard deviations (mm) by profile
type in experimental condition with no friction
Material
Profile
1
2
3
4
Actual
Break
Distance
7.5
15
22.5
30
Pre
M
24.7
24.3
30.8
31.2
Feedback
SD
12.7
9.6
6.8
5.3
M
13.4
14.9
23.8
NA
SD
8.9
2.8
3.5
NA
Post
M
15.1
15.1
24.2
27.5
Transfer
SD
9.5
2.2
4.7
3.8
M
13.4
14.4
24.5
27.5
SD
10.1
3.6
5.7
2.2
Table 2b
Average break point distance estimate means and standard deviations (mm) by profile
type in experimental condition with low friction (1.5N)
Material
Profile
1
2
3
4
Actual
Break
Distance
7.5
15
22.5
30
Pre
M
22.5
21.5
28.1
28.8
Feedback
SD
12.8
8.3
7.6
6.9
M
10.1
14.3
22.4
NA
29
SD
5.4
2.9
2.9
NA
Post
M
8.6
13.5
21
26.1
Transfer
SD
3.2
1.9
4.6
4.8
M
11.9
13.8
22.6
26.5
SD
8.8
2.3
4.4
2.4
Table 2c
Average break point distance estimate means and standard deviations (mm) by profile
type in experimental condition with high friction (3N)
Material
Profile
1
2
3
4
Actual
Break
Distance
7.5
15
22.5
30
Pre
M
22.7
21.3
28.2
29
Feedback
SD
12.6
8.7
8.2
7.1
M
NA
NA
NA
NA
SD
NA
NA
NA
NA
Post
M
8.2
13.3
19.8
23
Transfer
SD
3
3.3
4.6
7.9
M
9.5
13.3
20
24.8
SD
6.7
0.8
5.6
4.3
Figure 9. Average pretest break point estimates as a function of actual break point for all
participants in the experimental condition.
30
Figure 10. Average posttest break point estimates as a function of actual break point for
all participants in the experimental condition.
Means and standard deviations of distance are displayed by material type in the
control condition in Tables 3a, 3b, and 3c. Break point estimates from the pretest and
posttest, averaged across all participants in the experimental group are also displayed in
Figures 13 and 14.
31
Table 3a
Average break point distance estimate means and standard deviations (mm) by profile
type in control condition with no friction
Material
Profile
1
2
3
4
Actual
Break
Distance
7.5
15
22.5
30
Pre
M
25.6
21.3
28.9
31
Feedback
SD
11.4
8
7
4.6
M
20.1
21.7
28.7
NA
SD
10.9
7.6
5.5
NA
Post
M
20.7
20.9
27.8
29.7
Transfer
SD
10.2
7.2
5.8
3.8
M
6.8
13.4
20.6
26.2
SD
0.5
0.8
2.1
3
Table 3b
Average break point distance estimate means and standard deviations (mm) by profile
type in control condition with low friction (1.5N)
Material
Profile
1
2
3
4
Actual
Break
Distance
7.5
15
22.5
30
Pre
M
20.6
20
27.8
28.6
Feedback
SD
11.1
7.4
6.4
6.6
M
19.2
19.8
27.1
NA
SD
10
7.1
6
NA
Post
M
16.6
18.7
26.2
28.1
Transfer
SD
9.6
5.8
5.8
5.4
M
6.8
13.1
19.8
25.3
SD
0.6
1.5
3.1
3.9
Table 3c
Average break point distance estimate means and standard deviations (mm) by profile
type in control condition with high friction (3N)
Material
Profile
1
2
3
4
Actual
Break
Distance
7.5
15
22.5
30
Pre
M
18.6
18.7
24.4
25.2
Feedback
SD
10.6
7.9
7.6
8.3
M
NA
NA
NA
NA
32
SD
NA
NA
NA
NA
Post
M
15.2
17.5
21.8
25.1
Transfer
SD
9.4
7.1
7.9
6.7
M
6.6
13
18.9
22.5
SD
1
1
3.8
6
Figure 11. Average pretest break point estimates as a function of actual break point for
all participants in the control condition.
33
Figure 12. Average posttest break point estimates as a function of actual break point for
all participants in the control condition.
Simple regression models were used to determine the slopes and intercepts of the
functions predicting indicated distance from actual break point distance for each
participant and for each experimental phase. Then, they are used for the comparisons of
the contributors to perceptual estimates of distance of actual target distance and actual
force. Slopes, intercepts, and R2 values for both metrics for each participant across
34
phases are displayed in Tables 4a, 4b, 4c, and 4d for those in the experimental condition,
and Tables 5a, 5b, 5c, and 5d for those in the control condition. Perfect performance
estimating break point would result in a R2 = 1, slope = 1, and intercept = 0 for actual
distance and R2 = 0 for actual force. Slopes and intercepts given by regression techniques
are more useful than other descriptive statistics such as session means and signed error
because they describe the function that takes you from the actual target distances to the
perceived target distances. Trials in which participants broke the material were excluded
from analyses in the ToT Task, presented in Tables 4d and 5d, as the sudden decrease of
reactionary force would cause the participant’s estimate to fall close to or at the end of the physical limitations of the simulator.
35
Table 4a
Average R2, Slope, and Intercepts of Simple Regressions Predicting Break Point Estimates from Actual Break Point during the
Pretest Phase for Each Participant in the Experimental Condition
Participant
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
24
25
31
Avg
2
R
.082
.594
.168
.434
.000
.207
.116
.897
.162
.209
.249
.148
.262
.984
.843
.000
.066
.392
.303
.991
.981
.25
.008
.072
.908
.373
Friction 1
Slope
-.09
.65
.28
.43
.007
-.18
-.13
1.02
.35
-.4
.49
-.28
.5
.92
.7
.004
-.17
.38
.49
.89
.91
.5
.04
.16
1.08
.34
Intercept
39.2
18.8
27.4
22.2
32.8
39.2
38.7
-.2
14.7
38.4
16.5
29.8
16.6
.8
8
37.9
37.3
25
17.3
0
.9
9.8
34.5
29.5
-.8
21.4
36
2
R
.038
.184
.000
.408
.817
.108
.088
.686
.983
.478
.109
.019
.104
.976
.979
.002
.412
.425
.389
.994
.978
.998
.148
.000
.925
.450
Friction 2
Slope
.07
-.28
.004
-.82
.73
.25
.17
.69
.84
-.25
.21
.1
.28
.9
.9
.02
.57
.56
.5
.89
.85
1.02
-.28
.01
1.08
.36
Intercept
35.2
34.4
34.6
36.07
5.9
21.6
32.5
5
1.3
38.4
22.7
25.2
10.1
1.3
.9
33.2
10.5
19.5
21
1.2
1.2
-.3
37.4
29.7
-.9
18.3
2
R
.009
.159
.014
.099
.324
.407
.003
.963
.794
.345
.000
.42
.464
.992
.616
.223
.07
.243
.257
.96
.416
.987
.001
.005
.614
.375
Friction 3
Slope
.04
-.44
-.04
-.25
-.2
.41
-.02
.8
.58
-.26
-.001
.57
.59
.95
.7
-.17
.27
.49
.49
.86
.47
1
.009
-.05
.62
.3
Intercept
35.5
28.4
35.5
29.1
40.9
15.2
35.8
1.3
6.4
39.7
25.7
14.8
4.6
-.2
2
37.8
21.4
18.4
13.2
2.2
11.1
-.4
35.4
26.5
6.3
19.5
Table 4b
Average R2, Slope, and Intercepts of Simple Regressions Predicting Break Point Estimates from Actual Break Point during the
Feedback Phase for Each Participant in the Experimental Condition
Participant
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
24
25
31
Avg
2
R
.004
.787
.002
.002
.768
.89
.355
.724
.165
.957
.457
.458
.973
.995
.677
.893
.937
.87
.064
.136
.99
.981
.335
.242
.882
.582
Friction 1
Slope
-.06
.81
-.06
.04
.71
1
.56
.84
.42
1.1
.56
.49
1
1.03
.69
.75
.95
.99
.36
.54
1.04
.95
.59
.79
1
.68
Intercept
27.7
4.4
28
21
6.1
.6
8.3
3.1
9.3
-1
7.9
9.5
-.67
-.77
7.1
4.16
.52
1.59
15.05
9.5
-.76
1.07
7.71
6.84
0.5
7.1
37
2
R
.459
.591
.864
.517
.331
.975
.984
.967
.62
.055
.871
.887
.931
.98
.962
.44
.89
.859
.089
.975
.981
.991
.142
.983
.957
.732
Friction 2
Slope
.45
.62
.91
.84
.48
.97
1
1
.67
.33
.92
.98
.97
.95
1.03
.81
.88
1.09
.38
.92
.89
.98
.39
1
1.1
.82
Intercept
11.9
7.9
3.5
3.3
9.1
.7
-1.1
-.57
3
12.3
2
1.5
-.6
.07
-1.87
3.65
1.5
.43
12.42
.69
1.19
.22
10.4
-.31
-1.5
3.2
Table 4c
Average R2, Slope, and Intercepts of Simple Regressions Predicting Break Point Estimates from Actual Break Point during the
Posttest Phase for Each Participant in the Experimental Condition
Participant
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
24
25
31
Avg
2
R
.712
.486
.131
.18
.208
.278
.292
.031
.811
.964
.969
.052
.108
.994
.997
.991
.904
.518
.213
.998
.995
.99
.749
.211
.995
.591
Friction 1
Slope
.63
.55
.29
.24
.37
.38
.53
.14
.66
.89
.8
.19
.28
.92
.92
.84
.94
.55
.43
.93
.94
.95
.67
.42
.99
.62
Intercept
7.32
12.39
25.74
11.87
16.5
14.37
7.6
21.4
4.31
2.33
3.41
20.58
19.05
.58
.35
2.67
2.64
7.53
15.21
.45
.2
1.3
10.05
14.79
-.12
8.9
38
2
R
.083
.971
.877
.194
.858
.902
.555
.984
.997
.764
.493
.991
.988
.995
.991
.39
.989
.985
.961
.995
.993
.995
.824
.996
.995
.831
Friction 2
Slope
.17
.9
.81
.36
.7
.7
.74
.85
.86
.88
.63
.876
.92
.94
.93
.53
.86
.9
2.3
.93
.92
.93
.85
.95
.96
.86
Intercept
8.19
1.09
6
12.11
5.37
5.5
.16
.73
1.53
-.12
3.65
1.71
1.08
-.6
-.1
3.55
.95
.89
.91
0
.08
1.48
1.88
-.04
-.33
2.23
2
R
.203
.32
.084
.731
.809
.923
.976
.938
.117
.989
.607
.712
.995
.977
.955
.002
.996
.599
.902
.993
.683
.992
.659
.982
.95
.724
Friction 3
Slope
-.18
.58
.25
.76
.89
.69
.86
.82
.23
.89
.6
.68
.9
.97
.79
.03
.904
.7
.84
.88
.64
.94
.42
.96
.88
.68
Intercept
11.64
1.16
8.05
5.07
-1.35
5.06
.45
.56
8.73
1.32
3.32
7.22
.23
-1.95
.98
10.09
.433
3.22
3.14
.68
1.98
.72
11.9
.56
1.4
3.38
Table 4d
Average R2, Slope, and Intercepts of Simple Regressions Predicting Break Point Estimates from Actual Break Point during the
ToT Phase for Each Participant in the Experimental Condition
Participant
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
24
25
31
Avg
2
R
.997
.991
.991
.359
1.0
.994
.932
.995
.991
.995
.993
.994
.997
.997
.990
.997
.996
.995
.999
.992
.996
.999
.988
.994
.998
.967
Friction 1
Slope
.92
.94
.95
.36
.95
.95
.88
.9
.9
.92
.91
.95
.96
.93
.92
.93
.93
.93
.95
.91
.96
.99
.91
.95
1
.91
Intercept
-.01
.03
-.1
8
-.3
.1
.3
.01
.7
.3
.5
.3
.2
.4
-.4
.3
.2
-.1
-.08
.35
-.5
-.7
.5
-.5
-.6
.4
39
2
R
.991
.981
.995
.9
.985
.997
.995
.987
.991
.962
.999
.996
.993
.996
.995
.858
.992
.620
.998
.995
.994
.996
.987
.985
.994
.967
Friction 2
Slope
.94
.85
.9
.8
.9
.95
.87
.9
.84
.79
.94
.9
.92
.93
.88
.76
.83
.7
.92
.89
.9
.96
.86
.92
.96
.88
Intercept
-.4
1
.7
1.3
.2
.04
.7
-.1
1.2
1.8
-.2
.6
-.03
.2
.3
1.4
.6
3.1
.4
.2
.2
-.2
.1
-1.1
-.5
.5
2
R
.733
.991
.275
.432
.934
.918
.991
.834
.993
.994
.983
.943
.984
.996
.985
.983
.624
.990
.997
.988
.971
.998
.997
.978
.998
.900
Friction 3
Slope
.78
.84
.51
.33
.8
.75
.89
.68
.78
.92
.92
.87
.88
.97
.77
.86
.6
.89
.93
.86
.79
.98
.89
.84
.93
.81
Intercept
.3
1
6
6.5
-.2
2.6
.6
2.6
1.6
.01
.07
.4
.5
-1
1.1
.7
4.5
.6
.2
7
1.3
-.8
.07
.7
.04
1.5
Table 5a
Average R2, Slope, and Intercepts of Simple Regressions Predicting Break Point Estimates from Actual Break Point during the
Pretest Phase for Each Participant in the Control Condition
Participant
23
26
27
28
29
30
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
48
49
50
52
Avg
2
R
.041
.003
.051
.018
.110
.025
.946
.617
.164
.001
.024
.882
.077
.150
.784
.247
.616
.053
.334
.950
.154
.106
.101
.006
.769
.289
Friction 1
Slope
.1
-.05
.08
-.13
-.11
-.1
.9
.59
.42
-.02
.12
.89
.35
.38
.63
.36
.79
.13
.63
.87
.25
.24
.2
.05
.73
.33
Intercept
29.4
26.8
35
27.2
38
35.7
4.2
16.6
15.7
33.4
28.3
5.6
21.3
20.9
1.4
24.8
5.5
28.3
9.6
2.5
23.1
26.9
22.8
19.9
7.6
20.4
Friction 2
Slope
.08
.49
-.09
1.7
-.07
.53
.61
.58
.5
.47
.88
.66
1.1
.85
.15
.47
-.03
.18
.87
.94
.26
.03
.1
.05
.25
.45
2
R
.147
.309
.119
.966
.016
.298
.435
.428
.483
.468
.883
.452
.737
.666
.106
.275
.002
.118
.981
.874
.254
.037
.041
.008
.114
.369
40
Intercept
28
12.4
37
.86
31
11
7.2
14.3
13.1
11.9
3.1
13.5
0.9
6.7
7.3
19.1
35.8
25.8
2.1
-0.1
25.8
34.1
28.6
21.1
14.7
16.2
2
R
.088
.000
.001
.980
.464
.766
.008
.805
.965
.021
.094
.103
.980
.935
.002
.153
.272
.038
.922
.973
.001
.078
.290
.081
.553
.383
Friction 3
Slope
-.39
.02
-.01
.92
.48
.48
-.03
.85
.93
.09
.09
.29
.89
.84
.02
.19
.41
-.08
.85
.99
.02
-.06
.39
-.22
.74
35
Intercept
30.7
21.6
32.9
-0.2
13.3
9.4
12.8
5.9
1
15.1
28.5
6.6
0.8
4.2
6.4
26.3
17.6
30.1
2.5
0.3
25.9
34.2
21.4
28.7
6.1
15.3
Table 5b
Average R2, Slope, and Intercepts of Simple Regressions Predicting Break Point Estimates from Actual Break Point during the
Feedback Phase for Each Participant in the Control Condition
Participant
23
26
27
28
29
30
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
48
49
50
52
Avg
2
R
.000
.959
.191
.784
.821
.289
.923
.561
.318
.379
.985
.825
.004
.824
.871
.004
.294
.269
.776
.012
.871
.067
.167
.983
.138
.493
Friction 1
Slope
0
.85
.15
.74
1
.38
1.1
.57
9.91
.62
.98
1.13
.06
1
.84
.07
.32
.34
1.3
-.14
.68
-.1
.34
.95
.33
.94
2
Intercept
30.2
1.3
31.4
4.5
1.8
25.5
4.3
14
.6
9.9
-.8
.5
32
2.8
7.3
26.4
25.3
23.8
-3.1
33.1
16.1
36.5
25.2
-1.1
20.5
14.7
R
.003
.283
.125
.396
.499
.616
.861
.966
.462
.735
.818
.891
.102
.907
.843
.112
.01
.039
.762
.568
.309
.022
.269
.281
.515
.456
41
Friction 2
Slope
.04
.41
.11
.59
.7
.58
.84
.88
.61
.72
.77
1.1
.55
1
.75
.35
27.2
.09
1.2
.84
.448
.07
.47
-.58
.7
1.62
Intercept
29.8
9.5
33.5
5.8
8.8
21.9
8
2.9
9.8
4.4
1.8
2.4
13.5
.2
6.2
23.1
.06
29.4
-1.9
7.6
9.6
30.9
21.1
31.9
14.1
13
Table 5c
Average R2, Slope, and Intercepts of Simple Regressions Predicting Break Point Estimates from Actual Break Point during the
Posttest Phase for Each Participant in the Control Condition
Participant
23
26
27
28
29
30
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
48
49
50
52
Avg
2
R
.077
.858
.135
.056
.309
.252
.774
.961
.359
.903
.553
.408
.219
.295
.98
.000
.011
.528
.334
.327
.262
.002
.390
.994
.516
.420
Friction 1
Slope
.17
.74
.1
.15
.5
.23
.78
.92
.45
.91
.69
.45
-.14
.54
.79
.001
.04
.22
.62
.45
.41
-.02
.38
.91
.7
.44
Intercept
26.2
3.8
27.8
18.6
15.1
25.1
10.8
3.3
13.5
1.8
7.6
21.6
36.2
11.2
8.5
26.7
29.7
29.1
9.5
19
11.5
32.3
22.3
-1
6.6
16.7
Friction 2
Slope
-.09
.84
.07
.91
.94
.02
1
.96
.62
.83
.89
.96
.01
.14
.88
.54
.38
.11
.93
.73
.05
.11
.19
.87
.11
.52
2
R
.014
.988
.1
.994
.851
.001
.93
.968
.899
.885
.986
.867
.000
.984
.955
.416
.624
.056
.992
.792
.01
.106
.059
.99
.030
.58
42
Intercept
26.8
.5
24.7
.1
2.1
24.1
7
1.2
6.2
.2
.2
6.3
28.4
1
6.2
17.7
21.5
31.5
-.4
10.2
19.3
27.1
22.9
-.7
18.5
12.1
2
R
.026
.993
.07
.861
.215
.118
.967
.958
.084
.794
.992
.014
.994
.410
.577
.392
.56
.009
.985
.646
.128
.063
.482
.940
.55
.513
Friction 3
Slope
.12
.89
-.11
.69
.41
-.25
.92
.96
.17
.59
.85
-.099
.92
.79
.84
.5
.49
-.04
.97
.58
-.11
-.06
.37
.73
.24
.45
Intercept
24.7
-.3
31.5
2.3
9.1
25.6
5.9
-.89
10.2
6.8
1
18.3
.4
-1.9
4.7
21.5
11.4
32.1
-1.2
2.5
15.8
32.8
16.1
1.1
13.5
11.3
Table 5d
Average R2, Slope, and Intercepts of Simple Regressions Predicting Break Point Estimates from Actual Break Point during the
ToT Phase for Each Participant in the Control Condition
Participant
23
26
27
28
29
30
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
48
49
50
52
Avg
2
R
.996
.992
.957
.987
.977
.996
.995
.991
.997
.837
.995
.991
.996
.997
.967
.833
.997
.991
.994
.991
.464
.995
.992
.994
.988
.956
Friction 1
Slope
.93
.91
.77
.91
.88
.94
.93
.89
.86
.66
.93
.91
.91
.93
.92
.58
.98
1.09
.87
.91
.55
.89
.83
.94
.82
.87
Intercept
.2
-.6
2.5
.2
.5
-.3
.3
1.7
.9
1
.1
.7
.3
.3
-1.2
4.9
-.7
-3.9
.6
-.1
4.1
.3
1.3
-.8
.9
0.5
Friction 2
Slope
.94
.8
.62
.84
.83
.9
.89
.96
.88
.35
.9
.87
.93
.992
.87
.85
.94
.88
.79
.94
.62
.88
.71
.83
.77
.83
2
R
.993
.991
.797
.976
.884
.995
.995
.996
.994
.300
.986
.935
.991
.999
.981
.604
.994
.982
.947
.996
.754
.986
.814
.985
.939
.913
43
Intercept
-.3
1
3.2
1.3
0
.4
.4
-.5
.6
3.7
.3
.1
-.2
-.5
.9
-2.3
.1
.7
.7
-.2
2.6
.7
3
.1
1.3
.7
2
R
.382
.968
.294
.975
.993
.978
.990
.988
.978
.486
.990
.987
.992
.995
.320
.490
.698
.813
.978
.358
.759
.988
.715
.974
.539
.785
Friction 3
Slope
.43
.82
.22
.86
.91
.9
.9
.87
.8
.31
.9
.92
.91
.92
.45
.36
.51
.6
.97
.49
.5
.9
.65
.76
.56
.7
Intercept
6.7
.9
7.3
.6
-.06
.4
.4
.5
1.9
3.8
-.3
-.2
.5
.2
3.7
6.4
7.8
4.3
-2.4
3.9
5
.2
2.9
.3
3.5
2.3
Comparing Experimental & Control Conditions
Pretest. Examination of Tables 4a-d and 5a-d reveals that across friction levels,
the R2 values did not differ between the experimental and control conditions during the
pretest. A 3x2 mixed ANOVA using the R2 values from both conditions in the pretest
confirmed that there was no statistically significant difference in R2 values during pretest
performance, F(1, 48) = .413, p = .523. There also were no significant effect in the
difference in R2 values of break point estimates for different friction levels F(2, 96) =
1.139, p = .319, or significant condition by friction interaction, F(2, 96) = .495, p = .585.
Similarly, slope values did not differ between the experimental and control
conditions in the pretest. A 3x2 mixed ANOVA using slope values from both conditions
in the pretest confirmed no significant differences during pretest performance, F(1, 48) =
.237, p = .629. There also was no effect of friction on break point estimates across
participants F(2, 96) = 1.24, p = .291, or significant condition by friction interaction, F(2,
96) = .426, p = .626.
Similar to findings from R2 values and slope, intercept values observed in simple
regressions of individual participant performance did not differ between the experimental
and control conditions in the pretest. A 3x2 mixed ANOVA using intercept values from
both conditions in the pretest confirmed no significant differences during pretest
performance, F(1, 48) = .538, p = .467. However, there was an effect of friction across
the different friction levels F(2, 96) = 3.915, p = .023, but no significant condition by
friction interaction, F(2, 96) = .613, p = .544. Follow up paired samples t-tests revealed
that there were no significant differences in intercept values of simple regressions of
44
break point estimates between friction levels during the pretest within the experimental
condition. However, the change in R2 values in the control condition for materials with no
friction (M = .591, SD = .38) compared to low friction of 1.5N (M = .831, SD = .27)
revealed a significant difference between the estimates in the two friction levels, t(24) = 2.557, p = .017.
Posttest. A 3x2 mixed ANOVA using the R2 values from both conditions in the
posttest phase revealed a significant effect of condition F(1, 48) = 7.97, p < .01 indicating
that those who experienced the intervening calibration session tended to produce break
point estimates more strongly based on the actual break point. There was also a main
effect of friction level F(2, 96) = 5.893, p < .01, but no condition by friction interaction
F(2, 96) = .234, p = .792. Follow up paired samples t-tests revealed that there were no
significant differences in R2 values of break point estimates between friction levels during
the posttest within the control condition. However, the increase in R2 values in the
experimental condition for materials with no friction (M = .591, SD = .38) compared to
low friction of 1.5N (M = .831, SD = .27) revealed a significant difference between the
estimates in the two friction levels, t(24) = -2.557, p = .017.
A 3x2 mixed ANOVA of the slope values observed during posttest performance
for each participant revealed a significant effect of condition F(1, 48) = 10.48, p < .01
indicating that those who experienced the intervening calibration session tended to
produce simple regressions with a slope closer to 1. There was also a main effect of
friction level F(2, 96) = 4.73, p < .05, but no condition by friction interaction F(2, 96) =
1.12, p = .329. Follow up paired samples t-tests revealed that there were no significant
45
differences in break point estimates between friction levels during the posttest within the
control condition. However, the increase in slope values in the experimental condition for
materials with no friction (M = .62, SD = .28) compared to low friction of 1.5N (M = .86,
SD = .36) revealed a significant difference between the estimates in the two friction
levels, t(24) = -2.618, p = .015, as did the increase in slope values from materials with
high friction of 3N (M = .67, SD = .3) to low friction of 1.5N (M = .86, SD = .36), t(24) =
2.487, p = .02.
A 3x2 mixed ANOVA of the intercept values observed during posttest
performance for each participant revealed a significant effect of condition F(1, 48) =
15.757, p < .001 indicating that those who experienced the intervening calibration session
tended to produce simple regressions with an intercept closer to 0. There was also a main
effect of friction level F(2, 96) = 18.667, p < .001, but no condition by friction interaction
F(2, 96) = .628, p = .536. Follow up paired samples t-tests revealed significant
differences in break point estimates between friction levels during the posttest within the
control condition between materials with no friction (M = 16.7, SD = 10.6) and low
friction of 1.5N (M = 12.1, SD = 11.3) t(24) = 3.367, p = .003, as well as no friction (M =
16.7, SD = 10.6) and high friction of 3 N (M = 11.3, SD = 11.3) t(24) = 2.871, p = .008.
Within the experimental condition, intercept values also differed between materials with
no friction (M = 8.9, SD = 7.8) and low friction of 1.5N (M = 2.2, SD = 3.1) t(24) =
4.494, p < .0001 , and between materials with no friction (M = 8.9, SD = 7.8) and high
friction of 3N (M = 3.4, SD = 4), t(24) = 3.347, p = .003. The R2 values and the slopes of
the simple regressions tended to increase in the experimental group compared to the
46
control group, moving more closely to 1.0, and the intercepts decreased, moving more
closely to 0.
Comparing Pretest and Posttest Phases
Control Condition. Examination of Tables 4a-d and 5a-d reveals that across
friction levels, the R2 values differed across phases in the control condition. A 3x2
repeated measures ANOVA using the R2 values from both phases in the control condition
confirmed that the R2 values during pretest performance (M = .347, SE = .05) were
significantly lower than during posttest performance (M = .504, SE = .062) F(1, 24) =
5.536, p < .05. There was, however, no significant effect in the difference in R2 values of
break point estimates for different friction levels F(2, 48) = 1.888, p = .162, or significant
phase by friction interaction, F(2, 48) = .434, p = .651.
Slope values did not differ across phases in the control condition. A 3x2 repeated
measures ANOVA using slope values from both phases in the control condition
confirmed no significant differences across phase, F(1, 24) = 1.564, p = .223. There also
was no effect of friction on break point estimates across participants F(2, 48) = 1.24, p =
.291, or significant phase by friction interaction, F(2, 48) = 1.637, p = .205.
Similar to findings from slope values, intercept values observed in simple
regressions of individual participant performance did not differ across phases in the
control condition. A 3x2 repeated measures ANOVA using intercept values from both
phases in the control condition confirmed no significant differences across phase, F(1,
24) = 4.228, p = .051. However, there was an effect of friction across the different
friction levels F(2, 48) = 6.904, p < .05, but no significant phase by friction interaction,
47
F(2, 48) = .01, p = .99. Follow up paired samples t-tests revealed that there were
significant differences in intercept values of simple regressions of break point estimates
between friction levels of no friction (M = 18.55, SD = 10.93) and low friction of 1.5N
(M = 14.16, SD = 11.59), t(49) = 3.005, p < .05. There were also significant differences
in estimates between materials with no friction (M = 18.55, SD = 10.93) and high friction
of 3N (M = 13.3, SD = 11.62), t(49) = 3.556, p < .05. However, there were no differences
in estimates between materials with low friction (M = 14.16, SD = 11.59) and high
friction (M = 13.3, SD = 11.62), t(49) = .732, p = .468.
Experimental Condition. A 3x2 repeated measures ANOVA using the R2 values
from both phases in the experimental condition confirmed that the R2 values during
pretest performance (M = .399, SE = .065) were significantly lower than during posttest
performance (M = .715, SE = .041) F(2, 24) = 41.081, p < .05. There was also a
significant effect for different friction levels F(2, 48) = 4.779, p < .05, but no significant
phase by friction interaction, F(2, 48) = 1.26, p = .293. Follow-up paired samples t-tests
revealed significant differences between estimates of break points on materials with no
friction (M = .482, SD = .38) and low friction of 1.5N (M = .64, SD = .39), t(49) = 2.784, p < .05, as well as between materials with low friction of 1.5N (M = .64, SD = .39)
and high friction of 3N (M = .55, SD = .38), t(49) = 2.21, p < .05.
Slope values also differed across phases in the control condition. A 3x2 repeated
measures ANOVA using slope values from both phases in the experimental condition
revealed that slopes were steeper in the posttest (M = .33, SE = .08) than the pretest (M =
.72, SD = .04), F(1, 24) = 33.608, p < .05. There was no effect of friction on break point
48
estimates across participants F(2, 48) = .282, p = .052, or significant phase by friction
interaction, F(2, 48) = 1.827, p = .172.
A 3x2 repeated measures ANOVA using intercept values from both phases in the
experimental condition confirmed intercepts were significantly lower in the posttest (M =
4.84, SD = .74) than the pretest (M = 19.71, SD = 1.81), F(1, 24) = 39.024, p < .05. There
was also an effect of friction across the different friction levels F(2, 48) = 10.331, p <
.05, but no significant phase by friction interaction, F(2, 48) = 1.508, p = .232. Follow up
paired samples t-tests revealed that there were significant differences in intercept values
of simple regressions of break point estimates between friction levels of no friction (M =
15.14, SD = 13.05) and low friction of 1.5N (M = 10.27, SD = 13.27), t(49) = 3.773, p <
.05. There were also significant differences in estimates between materials with no
friction (M = 15.14, SD = 13.05) and high friction of 3N (M = 11.42, SD = 13.21), t(49)
= 3.012, p < .05. However, there were no differences in estimates between materials with
low friction (M = 10.27, SD = 13.27) and high friction (M = 11.42, SD = 13.21), t(49) = 1.217, p = .229.
In short, the results revealed an increase in the R2 values and improvements in
both slope and intercept, indicating a calibration effect that is characterized by an
improved scaling of the estimates to the actual target break point distances. Similar
improvements were seen in the control condition, but with smaller F-values than seen in
the experimental condition. Next, multilevel modeling techniques were used to determine
if the slopes and intercepts differed between the pretest and posttest sessions within each
of the conditions.
49
Multilevel Modeling
Due to the repeated measurements of the same participants, the trials of estimates
of break points are nested, or grouped, within participants. Variation in estimates may be
affected by individual differences, which affect performance throughout all of a
participant’s estimates. Nested data structures violate assumptions for performing an ordinary least squares regression or repeated-measures ANOVA (Bickel, 2007).
Therefore, multilevel modeling (MLM) was used to analyze the repeated measures model
in order to account for the nesting, or grouping, of trials within participants, which allows
for measurement occasions to be correlated. MLM accounts for correlated measurements
by estimating error separately for measurement occasions, or trials, and for individuals.
In the following MLM analysis, a two-level hierarchical structure is employed,
with each estimation trial defined as “Level 1” and each individual participant defined as
“Level 2,” allowing for the exploration of trial-level predictors, such as phase (pretest and
posttest) and friction levels, and person-level predictors such as condition assigned
(experimental versus control). The models employed were specified as:
𝐸𝑠𝑡𝑖𝑚𝑎𝑡𝑒𝑑 𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 = 𝛽 + 𝛽 (𝐴𝑐𝑡𝑢𝑎𝑙 𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒) + 𝛽 (𝐴𝑐𝑡𝑢𝑎𝑙 𝐹𝑜𝑟𝑐𝑒) +
𝛽 (𝑃ℎ𝑎𝑠𝑒) + 𝛽 (𝐹𝑟𝑖𝑐𝑡𝑖𝑜𝑛)
where 𝐸𝑠𝑡𝑖𝑚𝑎𝑡𝑒𝑑 𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 , participant i’s estimated break point at trial t, is a function
of an individual specific intercept parameter, 𝛽 ; participant-specific slope parameters,
𝛽 , capturing actual break point distance, 𝛽 , capturing actual break point force, 𝛽 ,
representing which phase the estimate occurred in (pretest or posttest), and 𝐵 ,
representing the level of friction applied to the material. Models were used to analyze
50
performance for the 25 participants in the experimental condition (N = 1753 trials),
separately from the 25 participants in the control condition (N = 1760 trials). Separate
models were then used to directly compare the two conditions.
It was important to first use an intercepts-only model without predictors to
provide a baseline (or null) model for the individual-level dependent variable, estimated
distance, and to determine if MLM was appropriate (Bickel, 2014). Results from the
intercept-only model indicate an intraclass correlation (ICC) of .109 in the experimental
condition, and .233 in the control condition, meaning that 11% and 23% of the total
variance of break point estimates, respectively, resides between participants, and 89%
and 77%, respectively, resides within participants, which supports the mixed model
approach. It is recommended that MLM analyses be used if ICC values exceed 0.05
(Heck, Thomas, & Tabata, 2010).
Predictor variables were centered around the grand mean to allow for comparison
of parameter estimates across models with both level-1 and level-2 predictors and to
reduce possible effects of collinearity. Predictors were entered into the model
hierarchically to determine their unique contribution to the model (see Tables 6 & 7).
Comparing Phases. Two models were analyzed, one for participants in each
condition, experimental and control. Predictors were added to the model as fixed effects
and random effects one at a time. If a predictor did not result in a significant effect, it
was removed from the model. In the first model, estimated break point distance was
entered as a dependent variable. Actual break point distance was entered as a fixed effect
to determine the linear trend of break point estimates across actual break points and also
51
entered as a random effect to determine the random error term. The final model included
all predictors that resulted in significant fixed or random effects. All predictors produced
significant fixed effects, and two produced significant random effects (actual distance and
friction) in each model.
2
Experimental Condition. Although R is not directly reported in MLM analyses
2
by statistical software, a pseudo- R was calculated, which represents the increase in
explained variance (and thus the reduction in residual variance) contributed by the
addition of a particular predictor. As seen in Table 6, actual break point distance, actual
break point force, phase, and friction all resulted in significant fixed effects, with phase
2
resulting in the greatest R effect size of 24.9% reduction in error variance, followed by
actual break point distance, 22.1%, friction, 4.3%, and actual force, 2.1%. As seen in
Figures 13 and 14 and Tables 4a & 4c, break point estimates for all friction levels became
more accurate in the posttest, with slopes becoming closer to 1 and intercepts becoming
closer to 0. Although performance significantly improved after training, it appears that
participants still tended to overestimate break point estimates for materials with low force
values (Materials 1 & 3) and no friction, and underestimate break point estimates with
high friction in the posttest. As evidenced in Figures 10 and 12 of posttest performance,
it’s possible the reactionary force in Material 1 fell below participants’ perceptual threshold when no friction was applied, as performance still lagged after training.
However, the introduction of friction, increasing the reactionary force as an operator
probes on the tissue, may cause fragile materials such as Material 1 to become above
perceptual threshold.
52
Table 6
Estimates of Fixed Effects and Standard Error (SE) in the Experimental Condition
Effect Intercept
Model 1 Model 2 Model 3 Model 4 Model 5 Estimate (SE) Estimate (SE) Estimate (SE) Estimate (SE) Estimate (SE) 22.025 (.684) 22.007 (.685) 22.007 (.686) 17.982 (.7004) 17.989 (.701)
Actual Distance
.514 (.052)*
Actual Force
.586 (.116)*
.586 (.053)*
.586 (.053)*
-1.357 (.219)* -1.367 (.179)* -1.367 (.173)*
Phase
8.018 (.32)*
Friction
8.012 (.309)*
-1.185 (.213)*
2
Change in Model R
-­-­
.221
*p < .05
53
.021
.249
.043
2
Calculations for Changes in Model R
Model 2 (unique effect of actual distance) o Model 1 residual – model 2 residual / model 1 residual
o (85.571 – 66.688) / 85.571 = .221
Model 3 (unique effect of actual force)
o Model 2 residual – model 3 residual / model 2 residual
o (66.688 – 65.256) / 66.688 = .021
Model 4 (unique effect of phase)
o Model 3 intercept – model 4 intercept / model 3 intercept
o (65.256 – 48.997) / 65.256 = .249
Model 5 (unique effect of friction)
o Model 4 intercept – model 5 intercept / model 4 intercept
o (48.997 – 46.886) / 48.997 = .043
Figure 13. Pretest break point estimates as a function of actual break point for all
participants in the experimental condition.
54
Figure 14. Posttest break point estimates as a function of actual break point for all
participants in the experimental condition.
Random effects. The intercept variance for those in the experimental condition
was estimated as 11.08, p < .05, so the estimate of the standard deviation is 3.33. This
suggests that there are unmeasured predictor variables for each participant that raise or
lower their performance. Individual participants will have intercepts that are within 3.33
mm higher or lower than the group average about 68% of the time, and within 6.66 mm
higher or lower 95% of the time.
Significant random effects of slope also occurred for predictors Actual Distance
and Friction (both p’s < .05). This suggests that the slope of actual distance predicting
55
estimated distance and friction predicting estimated distance varies for each participant.
Individual participant’s estimates predicted by actual distance will have slopes that vary within +0.24 mm of the group average 68% of the time, and within +0.48 mm of the
group average 95% of the time. Individual participant’s estimates predicted by friction will also have slopes that vary within +0.86 mm steeper than the group average 68% of
the time, and within +1.72 mm steeper than the group average 95% of the time. This
simply suggests that some participants produced more accurate break point estimates than
others, and some participants’ were more successful than others in ignoring variations in
friction level, with their estimates being less affected by the presence of friction.
Control Condition. As seen in Table 7, actual break point distance, actual break
point force, phase, and friction all resulted in significant fixed effects, similar to the
experimental condition. Although both conditions shared fixed and random effects,
differences can be seen in effect sizes. For participants in the control condition who did
2
not receive calibration training, actual break point distance resulted in the greatest R
effect size of 20.8% reduction in error variance, followed by friction, 8.6%, actual force,
3.5%, and phase, 1.7%. Although participants who did not receive training performed
better in the posttest than the pretest, likely due to practice effects, the effect of phase
accounted for only 1.7% of the variance compared to 24.9% in the experimental
condition. As seen in Figures 13 & 14 and Tables 4a & 4c, break point estimates for all
friction levels became somewhat more accurate in the posttest, with slopes becoming
closer to 1 and intercepts becoming closer to 0.
56
Although friction was a significant fixed effect in both experimental and control
2
group models, the R was relatively small at 8.6% and 4.3% respectively, accounting for
twice as much variance in the control group than the experimental group.
57
Table 7
Estimates of Fixed Effects and Standard Error (SE) in the Control Condition
Model 1 Model 2 Model 3 Model 4 Model 5 Effect Estimate (SE) Estimate (SE) Estimate (SE) Estimate (SE) Estimate (SE) Intercept
23.301 (.892)
Actual Distance
23.318 (.893) 23.32 (.895) 22.411 (.909) 22.417 (.909)
.427 (.02)*
Actual Force
.506 (.022)*
.506 (.022)*
.505 (.021)*
-1.468 (.184)* -1.468 (.183)* -1.461 (.175)*
Phase
1.818 (.327)* 1.815 (.313)*
Friction
-1.637 (.128)*
2
Change in Model R
-­-­
.208
*p < .05
58
.035
.017
.086
2
Calculations for Changes in Model R
Model 2 (unique effect of actual distance) o Model 1 residual – model 2 residual / model 1 residual
o (62.573 -­ 49.573) / 62.573 = .208
Model 3 (unique effect of actual force)
o Model 2 residual – model 3 residual / model 2 residual
o (49.573 – 47.848) / 49.573 = .035
Model 4 (unique effect of phase)
o Model 3 intercept – model 4 intercept / model 3 intercept
o (47.848 -­ 47.037) / 47.848 = .017
Model 5 (unique effect of friction)
o Model 4 intercept – model 5 intercept / model 4 intercept
o (47.037 -­ 43.001) / 47.037 = .086
Figure 15. Pretest break point estimates as a function of actual break point for all
participants in the control condition.
59
Figure 16. Posttest break point estimates as a function of actual break point for all
participants in the control condition.
Random Effects. The intercept variance for those in the control condition was
estimated as 19.54, p < .05, so the estimate of the standard deviation is 4.42. This
suggests that there are unmeasured predictor variables for each participant that raise or
lower their performance. Individual participants will have intercepts that are within 4.42
mm higher or lower than the group average about 68% of the time, and within 8.84 mm
higher or lower 95% of the time.
60
Similar to the experimental condition, significant random effects of slope also
occurred for predictors: Actual Distance and Friction, p’s < .05. This suggests that the slope of actual distance predicting estimated distance and friction predicting estimated
distance varies for each participant. Individual participant’s estimates predicted by actual distance will have slopes that vary within +0.24 mm of the group average 68% of the
time, and within +0.48 mm of the group average 95% of the time. Individual participant’s estimates predicted by friction will also have slopes that vary within +1.24 mm of the
group average 68% of the time, and within +2.48 mm of the group average 95% of the
time, which was greater variance for Friction than observed in the experimental
condition. This simply suggests that some participants produced more accurate break
point estimates than others, and some participants’ estimates were more affected by the presence of friction than others.
Comparing Conditions. Next, we consider if there was an effect of control
versus experimental condition on break point estimates. Two models were conducted,
one for pretest performance and one for posttest, with the same predictors as the final
model for examining effect of phase, but replaced effect of phase with effect of
Condition: Actual Distance, Actual Force, Friction, and Condition. As a Level 2
predictor, Condition was entered last into the model. There was a fixed effect of
Condition in the posttest, p < .05, which reduced the error variance by 22.6%. Condition
was not significant in the pretest, p = .489, indicating that there was no difference in
participant performance in each group before training.
61
Transfer of Training Phase
The Transfer of Training task was more realistic than the previous three phases, in
that materials would actually break if pushed past the simulated break point. As seen in
Tables 4d and 5d, performance was very similar between the experimental and control
conditions during the transfer of training phase, with those in the control group appearing
to underestimate break points more than those in the experimental group. Number of
breaks compared across friction level and material are displayed in Tables 8 & 9. Similar
to Long et al. (2014), the majority of breaks occurred when materials required the lowest
reactionary force before breaking (Materials 1 and 3). Breaks also appear to decrease as
friction increases.
An independent samples t-test shows that the number of times a participant broke
the tissue during the 36 trials in the experimental group (M = 4.64, SD = 3.09) was not
significantly different than the number of times a participant broke the tissue in the
control group (M = 4.36, SD = 2.8), t(48) = 0.336, p < .05. Although the number of
tissue breaks did not differ between the groups, it’s possible that combined with practice effects, those in the control group received enough haptic feedback within the first few
trials to judge tissue break points as well as those in the experimental group. Future
research should investigate different types of training (visual, haptic, or both) and the
number of trials needed for calibration to occur.
62
Table 8
Break Frequency Occurrence in the ToT Phase Across Material and Friction Level for
Participants in the Experimental Condition
Material
1
2
3
4
Total
No Friction
26/100
26%
2/73
2.7%
22/93
23.7%
0/73
0%
50/339 14.7%
Low Friction
27/96
28.1%
2/75
2.7%
16/86
18.6%
0/75
0%
45/332 13.6%
High Friction
17/90
18.9%
0/72
0%
6/75
8%
0/69
0%
23/306
7.5%
Total
70/286 24.5%
4/220
1.8%
44/254 17.3%
0/217
0%
118/977 12.1%
Table 9
Break Frequency Occurrence in the ToT Phase Across Material and Friction Level for
Participants in the Control Condition
Material
1
2
3
4
Total
No Friction
23/95
24.2%
5/75
6.7%
12/82
14.6%
1/72
1.4%
41/324 12.7%
Low Friction
21/90
23.3%
6/80
7.5%
14/87
16.1%
1/73
1.4%
42/330 12.7%
63
High Friction
24/94
25.5%
4/75
5.3%
4/79
5.1%
0/72
0%
32/320
10%
Total
68/279 24.4%
15/230
6.5%
30/248 12.1%
2/217
0.9%
115/974 11.8%
CHAPTER THREE
EXPERIMENT TWO
Experiment 2 tested if the phenomenon of detecting material break points
generalizes to a task other than pushing. The primary movements conducted during MIS
procedures include pushing, pulling, sweeping, and grasping (Singapogu et al., 2012b).
This experiment studied the generalizability of DTB to pulling motions, using the same
methodologies and procedures as in Experiment 1, except that participants will be
required to pull simulated tissues rather than push/probe them. If subjects can attune to
the invariant of DTB, they should be able to attune to DTB in other tasks such as pulling,
and results are expected to be the same as in Experiment 1.
Methods
Participants
23 university undergraduate students between the ages of 17 and 19 (M = 18.2,
SD = 0.6) Participated in Experiment 2 after providing informed consent, none of whom
had any experience practicing MIS. 12 were female and 11 were male. Participants
received course credit in exchange for their participation.
Materials, Apparatus and Procedures
All materials and procedures were the same as in Experiment 1, except subjects
were asked to pull simulated tissues using the laparoscopic device rather than push. The
feedback training phase was completed two to ten days after the pretest phase (M = 7, SD
= 2.2).
64
Results
Data were screened for outliers and for logging errors with the simulator. Due to
the restricted range of motion of the simulator, no trials exceeded a z-value of +3, so no
trials were excluded as outliers. However, two participants were excluded from the
analyses because they could not complete the task. Also, 15 pretest trials and 19 posttest
trials were not correctly recorded, with all values logged as 0, and were discarded.
Performance was assessed by analyzing displacement into the simulated material
via distance in millimeters. Means and standard deviations of distance are displayed by
material type and experimental phase in Tables 10a, 10b, and 10c. Break point estimates
from the pretest and posttest, averaged across all participants in the experimental group
are also displayed in Figures 17 and 18.
Table 10a
Average break point distance estimate means and standard deviations (mm) by profile
type and experimental phase with no friction
Material
Profile
1
2
3
4
Actual
Break
Distance
7.5
15
22.5
30
Pre
M
19.8
20.2
26.5
28.1
Feedback
SD
10.7
7.7
6.6
5.5
M
12.8
16.5
23.2
NA
65
SD
7.2
4.6
4.2
NA
Post
M
13.6
16.8
24.1
28.7
Transfer
SD
6.6
3.8
4.1
3.6
M
6.9
13.6
20.8
26.6
SD
0.4
0.7
1.3
2.6
Table 10b
Average break point distance estimate means and standard deviations (mm) by profile
type and experimental phase with low friction (1.5N)
Material
Profile
1
2
3
4
Actual
Break
Distance
7.5
15
22.5
30
Pre
M
16.2
21.1
22.6
24.1
Feedback
SD
9.9
9
9.5
8
M
11.5
15.1
21
NA
SD
6.7
4.5
6.2
NA
Post
M
12.4
15.2
22.6
26.1
Transfer
SD
6.7
3.1
4.4
6.2
M
6.8
13.2
20.3
25.5
SD
0.4
1.6
2.9
3.5
Table 10c
Average break point distance estimate means and standard deviations (mm) by profile
type and experimental phase with high friction (3N)
Material
Profile
1
2
3
4
Actual
Break
Distance
7.5
15
22.5
30
Pre
M
17.6
18.1
21.8
23
Feedback
SD
10.4
8.5
8.3
8.4
M
NA
NA
NA
NA
66
SD
NA
NA
NA
NA
Post
M
10.4
15
20.5
24.6
Transfer
SD
5.5
3.3
4.2
6.7
M
6.8
13.4
19.8
25.4
SD
0.5
1
3.2
4
Figure 17. Average pretest break point estimates as a function of actual break point for
all participants.
67
Figure 18. Average posttest break point estimates as a function of actual break point for
all participants.
Simple regression models were used to determine the slopes and intercepts of the
functions predicting indicated distance from actual break point distance for each
participant. Then, they are used for the comparisons of the contributors to perceptual
estimates of distance of actual target distance and actual force. Slopes, intercepts, and R2
68
values for both metrics for each participant across phases are displayed in Tables 11a,
11b, 11c, and 11d.
69
Table 11a
Average R2, Slope, and Intercepts of Simple Regressions Predicting Break Point Estimates from Actual Break Point during the
Pretest Phase for Each Participant
Participant
1
2
3
5
6
7
9
10
11
12
13
14
15
17
18
19
20
21
22
23
25
26
27
Avg
R2
.360
.748
.011
.005
.036
.043
.017
.374
.820
.877
.921
.950
.831
.202
.234
.114
.485
.004
.027
.915
.904
.741
.745
.451
Friction 1
Slope
.7
.49
.06
.06
.11
.11
-.1
.35
1
.68
1
.83
.67
.29
-.2
-.26
.37
-.02
-.08
.87
.86
1.1
.63
.41
Intercept
13.1
14.9
22.4
24.5
30.2
26.5
31.9
19.2
-.3
4.3
.4
.2
12.2
11.7
34.8
18.2
21.5
30.8
32.7
7.7
2.8
.2
5.3
15.9
70
R2
.434
.155
.653
.988
.179
.230
.65
.119
.860
.572
.000
.734
.959
.001
.073
.009
.046
.001
.292
.852
.062
.976
.003
.385
Friction 2
Slope
.83
-.31
.96
.9
.58
.26
.59
-.3
.67
.38
-.03
.73
.86
-.01
.15
.03
-.13
.02
-.31
.72
.26
.99
.03
.34
Intercept
10.4
20
.8
-.6
12.8
25
5.5
26.5
2.8
9.7
14.4
2
4.1
20.9
25
10
27.8
30.7
37.6
11.8
6.9
-.9
32.4
14.6
R2
.435
.055
.152
.118
.563
.281
.192
.463
.015
.457
.303
.492
.516
.273
.221
.033
.517
.057
.264
.316
.963
.047
.934
.333
Friction 3
Slope
.7
-.12
-.35
.4
.69
-.38
.26
.54
-.07
.54
.69
.41
.61
.28
-.23
-.08
.24
.15
-.41
.52
.77
.25
.86
.27
Intercept
9.6
13.3
19.7
17.7
4.5
32.6
9.5
8.7
30.1
6.2
6.7
3.3
8.9
6.7
31.9
11.6
24.8
25.1
36
10.5
4.8
22.2
1.4
15
Table 11b
Average R2, Slope, and Intercepts of Simple Regressions Predicting Break Point Estimates from Actual Break Point during the
Feedback Phase for Each Participant
Participant
1
2
3
5
6
7
9
10
11
12
13
14
15
17
18
19
20
21
22
23
25
26
27
Avg
R2
.814
.992
.209
.805
.673
.650
.774
.644
.062
.246
.999
.132
.917
.867
.634
.967
.158
.361
.106
.797
.998
.980
.067
.602
Friction 1
Slope
1.1
.93
.47
.81
.66
1
1.2
.69
.31
.49
1
.37
.87
.86
.98
.91
.4
.59
.38
.79
.98
1.08
-.18
.73
Intercept
2.4
.8
10.3
3.3
7.6
2.2
-1.3
3.6
16.6
10
-.3
15.6
1.8
3.9
2.7
.5
14.4
9.1
11.7
5
-.1
-.8
31.8
6.6
71
R2
.852
.008
.631
.898
.813
.964
.206
.107
.566
.340
.608
.124
.996
.066
.022
.773
.523
.964
.421
.780
.270
.976
.002
.518
Friction 2
Slope
.72
.12
.91
.88
.83
.94
.46
.44
1
.65
.76
.2
1
.23
.14
.9
.85
1.13
.49
.84
.59
.97
-.06
.65
Intercept
4.3
15.3
.5
2.1
1.8
1
7.4
6.8
4.1
7
2.2
10.9
-.2
14.3
14.3
.5
1.3
-2.2
6.8
3.9
7.2
.1
26.7
5.9
Table 11c
Average R2, Slope, and Intercepts of Simple Regressions Predicting Break Point Estimates from Actual Break Point during the
Posttest Phase for Each Participant
Participant
1
2
3
5
6
7
9
10
11
12
13
14
15
17
18
19
20
21
22
23
25
26
27
Avg
R2
.966
.994
.853
.980
.095
.817
.867
.222
.874
.988
.997
.898
.965
.919
.409
.387
.965
.383
.649
.974
.985
.993
.120
.752
Friction 1
Slope
.9
.97
.96
.97
.21
.81
.89
.39
.8
.87
.94
.77
.89
.74
.49
.32
.93
.33
.51
.83
.75
.92
.19
.71
Intercept
3.9
.3
5.4
-.1
22.2
7.8
5.8
11.5
6.2
2.7
.5
5
3.5
5.2
10.1
15.4
1.2
19.6
11.5
3.8
3.9
-.6
25.5
7.4
72
R2
.396
.981
.698
.982
.427
.568
.916
.035
.399
.614
.993
.245
.984
.747
.879
.314
.183
.797
.393
.982
.992
.985
.079
.634
Friction 2
Slope
.71
.93
.75
.31
.54
.53
.93
.16
.53
.69
.92
.3
.88
.44
.85
.41
.28
.66
.33
1.1
.9
.95
.18
.6
Intercept
3.2
.2
8
.9
11
10.4
4.5
10.7
11.3
4.4
.2
16.4
2
15.1
3.2
7.9
16.6
9.9
7.2
-3.1
.4
-.5
20.1
7
R2
.982
.990
.845
.993
.768
.919
.747
.021
.334
.340
.991
.000
.652
.042
.919
.716
.084
.030
.945
.467
.845
.994
.980
.635
Friction 3
Slope
.97
.95
.73
.95
.74
.83
.98
.1
.38
.37
.94
.01
.74
.14
.87
.65
-.23
.09
.74
.65
.93
.95
.86
.62
Intercept
.8
.7
6.7
-.8
4
4.9
-1.5
11.9
14.5
7.9
.87
20.4
1.8
15.8
2.7
6.4
23.7
12.2
2.4
3.2
-1
-.04
2
6.1
Table 11d
Average R2, Slope, and Intercepts of Simple Regressions Predicting Break Point Estimates from Actual Break Point during the
ToT Phase for Each Participant
Participant
1
2
3
5
6
7
9
10
11
12
13
14
15
17
18
19
20
21
22
23
25
26
27
Avg
R2
.995
.996
.994
.993
.861
.996
.962
.997
.996
.995
.994
.988
.997
.92
.986
.997
.992
.990
.926
.994
.977
.995
.991
.980
Friction 1
Slope
.97
.9
.94
.92
.58
.94
.92
1
.9
.93
.96
.92
.89
.8
.88
.92
.93
1
.74
.89
.87
.97
.91
.90
Intercept
-.3
.5
-.6
-.2
4
-.1
-.5
-.7
-.9
-.7
-.3
-.1
.3
2
.9
.3
-.1
-1.3
2.1
.8
.5
-.5
.1
.2
73
R2
.992
.994
.98
.995
.732
.997
.928
.892
.993
.845
.993
.985
.994
.665
.985
.991
.989
.749
.390
.992
.982
.997
.992
.915
Friction 2
Slope
.95
.95
.86
.9
.57
.97
.85
.74
.89
.83
.92
.86
.91
.68
.9
.87
.87
1
.41
.88
.87
.96
.86
.85
Intercept
.1
-.3
.7
-.1
5
-.3
.8
2.2
-.1
-.2
-.2
1
-.1
2.2
-.1
.4
.9
-5.8
5.8
.5
1.4
-.4
.3
.6
R2
.991
.996
.995
.927
.442
.986
.973
.847
.996
.993
.993
.796
.998
.572
.902
.988
.973
.494
.764
.996
.986
.999
.996
.896
Friction 3
Slope
.95
.94
.94
.87
.39
.9
.94
.81
.94
.9
.91
.58
.94
.52
.79
.83
.9
.7
.65
.9
.9
.95
.9
.83
Intercept
-.4
-.1
-.2
-.4
4
.5
-.7
1.7
-.4
-.2
.2
4.3
-.03
3.9
2.3
1.1
.6
2
4
.3
.8
-.1
-.4
1
Comparing Pretest and Posttest Phases
Examination of Tables 11a-d reveals that across friction levels, the R2 values did
differ across phases. A 3x2 repeated measures ANOVA using the R2 values from both
phases confirmed that the R2 values during pretest performance (M = .39, SE = .05) were
significantly lower than during posttest performance (M = .674, SE = .052) F(1, 22) =
21.465, p < .05. There was, however, no significant effect in the difference in R2 values
of break point estimates for different friction levels F(2, 44) = 1.529, p = .228, or
significant phase by friction interaction, F(2, 44) = .151, p = .86.
Slope values also differed across phases. A 3x2 repeated measures ANOVA using
slope values from both phases revealed significant differences in break point estimates
between the pretest (M = .343, SE = .065) and posttest (M = .652, SE = .049), F(1, 22) =
21.001, p < .05. There was, however, no effect of friction on break point estimates across
participants F(2, 44) = 1.627, p = .208, or significant phase by friction interaction, F(2,
44) = .198, p = .821.
Similar to findings from R2 and slope values, intercept values observed in simple
regressions of individual participant performance also varied across phase. A 3x2
repeated measures ANOVA using intercept values from both phases in the control
condition confirmed significant differences between pretest (M = 15.2, SE = 1.838) and
posttest (M = 6.809, SE = 1.121), F(1, 22) = 4.228, p = .051. There was no effect across
the different friction levels F(2, 44) = .225, p = .799, and no significant phase by friction
interaction, F(2, 44) = .135, p = .874.
74
In short, the results showed an improvement in break point estimates in the
posttest compared to the pretest, with R2 and slope values closer to 1 and intercept values
closer to 0, indicating a calibration effect that is characterized by an improved scaling of
the estimates to the actual target break point distances. Next, multilevel modeling
techniques were used to determine if phase and friction produced main effects while
accounting for variance between and within participants.
Multilevel Modeling
Models similar to those employed in Experiment 1were used to analyze
performance for the 23 participants in the experiment (N = 1622 trials). It was important
to first use an intercepts-only model without predictors to provide a baseline (or null)
model for the individual-level dependent variable, estimated distance, and to determine if
MLM was appropriate. Results from the intercept-only model indicate an intraclass
correlation (ICC) of .097, meaning that 9.7% of the total variance of break point
estimates, resides between participants, and 90.3% resides within participants, which
supports the mixed model approach. It is recommended that MLM analyses be used if
ICC values exceed 0.05 (Heck et al., 2010).
Predictor variables were centered around the grand mean to allow for comparison
of parameter estimates across models with both level-1 and level-2 predictors and to
reduce possible effects of collinearity. Predictors were entered into the model
hierarchically to determine their unique contribution to the model (see Table 12).
Comparing Pretest and Posttest. Predictors were added to the model as fixed
effects and random effects one at a time. As with the previous push gesture experiment,
75
all predictors produced significant fixed effects, and two produced significant random
effects (actual distance and friction). As seen in Table 12, actual break point distance,
actual break point force, phase, and friction all resulted in significant fixed effects, with
2
actual distance resulting in the greatest R effect size of 26.7% reduction in error
variance, followed by friction, 3.9%, phase, 3%, and actual force, 0.6%. As seen in
Figures 19 and 20 and Tables 11a & 11c, break point estimates for all friction levels
became more accurate in the posttest, with slopes becoming closer to 1 and intercepts
becoming closer to 0. Results were very similar to those in the push gesture experiment,
both exhibiting the same fixed and random effects, with similar effect sizes. However,
although performance significantly improved after training in the current experiment,
phase reduced the error variance by only 3%, rather than 24.9% as with the push gesture
experiment, which may be largely due to better pretest performance in the current
experiment. It also appears that participants did not tend to overestimate break point
estimates for materials with low force values (Materials 1 & 3) and no friction, as they
did in the push gesture experiment.
76
Table 12
Estimates of Fixed Effects and Standard Error (SE) in the Experimental Condition
Effect Intercept
Model 1 Model 2 Model 3 Model 4 Model 5 Estimate (SE) Estimate (SE) Estimate (SE) Estimate (SE) Estimate (SE) 20.399 (.592) 20.401 (.592) 20.403 (.593)
19.19 (.615)
19.181 (.614)
.499 (.047)*
.533 (.048)*
.534 (.048)*
.536 (.048)*
-.635 (.188)*
-.639 (.184)*
-.644 (.176)*
2.42 (.33)*
2.429 (.315)*
Actual Distance
Actual Force
Phase
Friction
-1.089 (.277)*
2
Change in Model R
-­-­
.267
*p < .05
77
.006
.030
.039
2
Calculations for Changes in Model R
Model 2 (unique effect of actual distance) o Model 1 residual – model 2 residual / model 1 residual
o (66.478 – 48.72) / 66.478 = .267
Model 3 (unique effect of actual force)
o Model 2 residual – model 3 residual / model 2 residual
o (48.72 – 48.42) / 48.72 = .006
Model 4 (unique effect of phase)
o Model 3 intercept – model 4 intercept / model 3 intercept
o (48.42 – 46.985) / 48.42 = .030
Model 5 (unique effect of friction)
o Model 4 intercept – model 5 intercept / model 4 intercept
o (46.985 – 45.157) / 46.985 = .039
Figure 19. Pretest break point estimates as a function of actual break point for all
participants.
78
Figure 20. Posttest break point estimates as a function of actual break point for all
participants.
Random effects. The intercept variance for those in the experimental condition
was estimated as 7.51, p < .05, so the estimate of the standard deviation is 2.74. This
suggests that there are unmeasured predictor variables for each participant that raise or
lower their performance. Individual participants will have intercepts that are within 2.74
mm higher or lower than the group average about 68% of the time, and within 5.48 mm
higher or lower 95% of the time.
79
Significant random effects of slope also occurred for predictors Actual Distance
and Friction, p’s < .05. This suggests that the slope of actual distance predicting estimated distance and friction predicting estimated distance varies for each participant.
Individual participant’s estimates predicted by actual distance will have slopes that vary within +0.2 mm of the group average 68% of the time, and within +0.41 mm of the group
average 95% of the time. Individual participant’s estimates predicted by friction will also
have slopes that vary within +1.17 mm of the group average 68% of the time, and within
+2.35 mm of the group average 95% of the time. This simply suggests that some
participants produced more accurate break point estimates than others before training,
and some participants’ estimates were more affected by the presence of friction than others.
Transfer of Training Phase
Number of breaks compared across friction level and material are displayed in
Table 13. Similar to Long et al. (2014) and Experiment 1, the majority of breaks
occurred when materials required the lowest reactionary force before breaking (Materials
1 and 3). However, breaks do not appear to decrease as friction increases as they did in
Experiment 1.
80
Table 13
Break Frequency Occurrence in the ToT Phase Across Material and Friction Level
Material
1
2
3
4
Total
No Friction
30/99
30.3%
7/82
8.5%
21/95
22.1%
1/73
1.4%
59/349 16.9%
Low Friction
21/91
23.1%
2/74
2.7%
10/82
12.2%
0/75
0%
33/322 10.2%
81
High Friction
32/96 33.3%
3/76
3.9%
12/85 14.1%
0/74
0%
47/331 14.2%
Total
83/286
29%
12/232
5.2%
43/262 16.4%
1/222
0.5%
139/1002 13.9%
CHAPTER FOUR
GENERAL DISCUSSION
The current experiments demonstrated the ability of novices to detect soft tissue
break points in the presence of friction, particularly in the context of a simulated MIS
task. Two experiments were conducted to investigate whether participants were sensitive
to DTB through applied force on a MIS tissue simulator. In the first experiment,
participants probed four simulated soft tissues at three levels of friction (no friction, low
friction, and high friction), where training was found to improve sensitivity to DTB
through attunement and calibration. In the second experiment, participants’ sensitivity to DTB was observed in a pulling task, another motion common during MIS procedures.
Hypothesis 1
It was hypothesized that participants are sensitive to DTB, which was supported
with simple regressions of pretest performance for materials without friction simulated
(Tables 4a, 5a, and 11a). Perfect performance would be represented as having R2 and
slope values of 1, and intercepts of 0. Simple regressions predicting break point
estimates from actual tissue break point produced positive R2 values of .373 and .289 for
the experimental and control conditions in Experiment 1, respectively, and .451 in
Experiment 2. Slope values were also positive values of .34 and .33 for the two groups in
Experiment 1 and .41 in Experiment 2. Intercept values were 21.4 and 20.4 for the two
groups in Experiment 1 and 15.9 in Experiment 2. These values indicate that participants
could complete the task, although slopes of 21.4, 20.4, and 15.9 indicate overestimation
of break points, especially for tissues that broke early.
82
Hypothesis 2
It was also hypothesized that novice participants are able to detect DTB with
varying levels of friction present, which was supported with simple regressions of
performance data in the pretest (Tables 4a, 5a, and 11a). Participants were able to
complete the task in the pretest, with no difference in performance between experimental
and control conditions. Multilevel modeling techniques revealed that the main contributor
to break point estimates was the actual break point distance location, which participants
must detect by utilizing the change in haptic force as distance into the tissue was
manipulated. Even as lower-order parameters, actual break point distance and force,
varied across materials, participants were successfully able to detect break points,
indicating sensitivity to DTB. Because friction components do not affect the change in
reactionary force as displacement into the tissue increases, it was hypothesized that
participants would successfully be able to attune to DTB in the presence of friction.
Although there was an effect of friction on break point estimates, this appears to
be due primarily to two trends. First, when 3N of friction, the highest level observed by a
trocar in actual laparoscopic surgery (van den Dobbelsteen, Schooleman, & Dankelman,
2007), was applied participants tended to underestimate tissue break point, particularly as
actual break point distance increased. In other words, when reactionary force was high
immediately upon probing, participants were hesitant to push as far into the simulated
tissue as when reactionary force was lower. Second, participants tended to overestimate
the break point of tissues when no friction was present, particularly when the actual break
point reactionary force was low. This result suggests that the presence of trocar friction
83
may have the positive effect of causing users to be more cautious, and thus accidently
breaking tissues less. Future research should investigate the possibility that this trend is
likely caused by these materials of low reactionary force falling below perceptual
threshold. Without perceiving contact with the tissue, break point estimates fell toward
the end of the tissue simulator, as participants searched for contact.
Hypothesis 3
Performance data in the posttest supports the third hypothesis, that sensitivity to
DTB is a skill that can be improved with training. Similar to Long et al. (2014),
participant break point estimates were significantly more accurate after attunement and
calibration to DTB during a brief training phase. These performance improvements were
observed across all four materials and three friction levels, even though training only
provided feedback for three of the materials and two of the friction levels, indicating
sensitivity to DTB rather than lower-order parameters such as actual break point force
and distance. . In Experiment 1, average R2 values of participant performance in the
posttest increased 58% for materials with no friction, 85% for those with low friction, and
93% for those with high friction compared to pretest performance, indicating that break
point estimates were more consistent and precise as they became more sensitive to the
mechanical information specifying DTB. Slopes of the simple regressions also
significantly improved in the posttest, approaching perfect performance of 1, and
intercepts significantly improved, approaching perfect performance of 0. Multilevel
modeling also revealed that posttest performance was significantly better for those in the
84
experimental condition than those in the control condition, who did not receive visual
feedback after each estimate in the feedback phase.
Similar to the pretest, there was an effect of friction on break point estimates after
training, reducing the error variance by 8.6% in the posttest compared to 4.3% in the
pretest. Although the effect size was greater in the posttest, this increase in effect size
was due to the improved performance despite friction, with the exception of the first
material. Materials 1 and 3 may have been below perceptual threshold when friction was
not present, with 90% and 77% of estimates in the pretest, respectively, and 81% and
64% of estimates in the posttest, respectively, exceeding the actual break point.
However, break point estimates become more accurate on Material 1 as friction was
added (see Figure 10). Because performance was much more accurate and consistent in
the posttest compared to the pretest, this effect of friction appears larger in the posttest.
Future research should investigate the possibility that friction actually causes fragile
materials with break points below perceptual threshold to become above threshold.
Unlike past research which has shown that friction can cause perceptual thresholds to
increase (Perreault & Cao, 2006), the current study suggests that friction does not appear
to be a variable that surgeons must overcome, but is something that surgeons can learn to
ignore with training while it may simultaneously increase sensitivity to DTB. Friction
may assist in the perception of other variables, such as reactionary force or distance,
required to attune to DTB. These differences may be due to the different methodologies
(training vs. no training), equipment (haptic simulator vs. real silicone materials), or
differences in the fragility of materials presented.
85
Similar to Pagano and Cabe (2003; Pagano & Donahue, 1999), participants were
successfully able to attune to and differentiate between useful haptic information and
non-specifying variables. Just as the human perceptual system had shown to accurately
perceive the length of a rod by attuning to the invariant of inertia, ignoring effects of
wielding in different media, novices in both Experiment 1 and 2 were successfully able to
ignore trocar friction, attuning to the invariant of DTB. However, there are times when
the additional muscular forces needed to overcome the added friction may have made it
easier to attune to useful information in the haptic array, which allowed participants in
the current study to make more accurate break point estimates.
Hypothesis 4
It was hypothesized that sensitivity to DTB would transfer to a task where the
participants must stop before the break point is reached, which was supported by
performance in the ToT phase of both experiments. Simple regressions predicting break
point estimates from actual break point (Tables 4d and 5d) for trials in which the break
point was not exceeded reveal that estimates were near perfect performance, with an
average R2 value of .945 across friction levels in the experimental condition and .885 in
the control condition of Experiment 1. In Experiment 2, R2 values were similarly high,
with an average of .930 across friction levels. Slope values were near perfect, with
average values of 0.87 across friction levels in the experimental condition and 0.8 in the
control condition of Experiment 1, and 0.86 in Experiment 2. Average intercept values
were 0.8 in the experimental condition and 1.2 in the control condition of Experiment 1,
and 0.6 in Experiment 2.
86
Participants only broke 12.1% of the tissues in the experimental condition, with
breaks decreasing with increased friction, and 11.8% of the tissues in the control
condition. Although participant estimates in the control condition appeared to be less
accurate than those in the experimental condition, they did not break more tissues, likely
due to practice effects and possible rapid calibration to DTB once haptic feedback was
received upon breaking materials. Those in Experiment 2 only broke 13.9% of the
tissues when pulling tissues. Similar to Long et al. (2014), breaks were most likely to
occur when materials required the lowest reactionary force before breaking.
Hypothesis 5
Finally, it was also hypothesized that DTB generalizes to other MIS motions or
tasks, such as pulling, which was supported by findings in Experiment 2, which
employed the same procedure and materials as Experiment 1 but required that
participants pull the simulated tissue rather than push to identify break point. Similar to
the first experiment, participant break point estimates were significantly more accurate
across all four materials and three friction levels after attunement and calibration to DTB
during a brief training phase that only provided feedback for three of the materials and
two of the friction levels. Performance in Experiment 2 revealed the same main effects
(actual distance, actual force, phase, and friction) as observed in Experiment 1, and the
same random effects (actual distance and friction).
Simple regressions of individual participant performance reveal that in
Experiment 2, as in Experiment 1, R2 values significantly improved 67% for materials
without friction, 65% for materials with low friction, and 91% for materials with high
87
friction after training, indicating that estimations became more precise and consistent.
Slope and intercept values also became more accurate, approaching perfect performance
of 1 and 0, respectively. However, unlike Experiment 1, Materials 1 and 3 did not appear
to fall below perceptual threshold when friction was not present. Although these may
still be difficult materials to detect break point because of the fragility, it’s possible that the muscular forces needed to move the tool, hand and arm against gravity during the
pulling motion, rather than with gravity as in the pushing motion, may have helped
participants perceive the low force values. Thus in Experiment 2, gravity may have
played the facilitating role served by friction in Experiment 1.
Conclusions
The current study was one of the first to examine how trocar friction affects soft
tissue break point estimates within the context of a simulated MIS task. With a brief 10minute training using our Core Haptic Skills Training Simulator, 18 to 20-year-old
undergraduate psychology students’ estimates of tissue break points significantly improved, even in the presence of friction. However, future research should continue to
investigate certain shortcomings of the current experiments. For example, the current
study only examined simulated tissues and simulated trocar friction. Future research
should investigate if the observed improved performances transfer to real biological
tissues and real trocars. Although the current simulated materials follow the exponential
stress-strain pattern as exhibited in actual soft tissues (Brouwer et al., 2001; Fung, 1993;
Rosen, et al., 2008; Yamada, 1970), it’s important to determine if the training generalizes
to actual surgical procedures and other stick and slip effects of the rubber trocar. If so,
88
surgical training programs should employ the current haptic training to reduce the
number of accidental breakage of healthy tissues during surgery. Future work should
also research other possible types of feedback during training, such as haptic feedback
alone, or a combination of visual and haptic feedback to maximize effects of attunement
and calibration. Additionally, research should further investigate perceptual threshold for
pushing motions and pulling motions and the possibility that the addition of friction can
cause these fragile tissues to become perceptible.
One other shortcoming of the current experiments is that all participants were
novice psychology students without any other surgical training. Past research, however,
indicates that while experienced surgeons have an increased ability to detect DTB, their
overall performance is very similar to that of the participant population employed in the
present study (Long et al., 2014). Future work should confirm if the current findings
generalize to actual surgeons, or if the improved performance resulting from the current
methods only apply to those with no previous experience. However, it was also found
that although surgeons demonstrated better performance than novices overall, some
surgeons may be better at detecting DTB than others. Additionally, other research has
demonstrated that receiving haptic feedback in a virtual training environment may be
especially critical during early training phases for psychomotor skill acquisition (Ström et
al., 2006).
Surprisingly, break point estimates in the current study became more accurate
when friction was present than when it was absent. Previous research on friction and soft
tissue break points in MIS has primarily assumed that friction is a hindrance, which
89
surgeons must learn to overcome, and which engineers must work to reduce (Perreault &
Cao, 2006; van den Dobbelsteen, Schooleman, & Dankelman, 2007). However, with
other variables controlled, such as laparoscopic tool angle and visual feedback, it’s possible that surgeons may be easily trained to attune to mechanical properties specifying
DTB, using friction to assist them when reactionary force values of fragile tissues fall
below perceptual threshold. Perhaps researchers should seek to identify an optimal level
of friction, as too much friction may cause underestimation of break points, and too little
may prevent fragile tissues from being haptically perceived. Minimizing preventable
damages and injuries during MIS is achievable goal, one which can be improved by
further research of mechanical information specifying DTB and by employing effective
training MIS haptic training programs.
90
REFERENCES
Altenhoff, B. M., Napieralski, P. E., Long, L. O., Bertrand, J. W., Pagano, C. C., Babu, S.
V., & Davis, T. A. (2012). Effects of calibration to visual and haptic feedback on
near-field depth perception in an immersive virtual environment. Proceedings of
the ACM Symposium on Applied Perception, 71-78.
Archer, S. B., Brown, D. W., Smith, C. D., Branum, G. D., & Hunter, J. G. (2001). Bile
duct injury during laparoscopic cholecystectomy: Results of a national survey.
Annals of Surgery, 234, 549-559.
Bergmann Tiest, W. M., & Kappers, A. M. L. (2009). Cues for haptic perception of
compliance. IEEE Transactions of haptics, 2(4), 189-199.
Bethea, B. T., Okamura, A. M., Kitagawa, M., Fitton, T. P., Cattaneo, S. M., Gott, V. L.,
Baumgartner, W. A., & Yuh, D. D. Application of haptic feedback to robotic
surgery. In Journal of Laparoendoscopic & Advanced Surgical Techniques, 14, 3,
(2004), 191-195.
Bickel, R. (2007). Multilevel analysis for applied research: It’s just regression! New
York: The Guilford Press.
Bingham, G. P. & Pagano, C. C. (1998). The necessity of a perception-action approach to
definite distance perception: Monocular distance perception to guide reaching.
Journal of Experimental Psychology, 24(1), 145-168.
Bliss, J. P., Tidwell, P. D., & Guest, M. A. (1997). The effectiveness of virtual reality for
administering spatial navigation training to firefighters. Presence: Teleoperators
and Virtual Environments, 6, 73–86.
Breedveld, P., & Wentink, M. (2001). Eye-hand coordination in laparoscopy – an
overview of experiments and supporting aids, Minimally Invasive Therapy &
Allied Technology, 10(3), 155-162.
Brouwer, I, Ustin, J., Bentley, L., Sherman, A., Dhruv, N. & Tendick, F. (2001).
Measuring in vivo animal soft tissue properties for haptic modeling in surgical
simulation. Studies in Health Technologies and Informatics, 81, 69-74.
Cabe, P. A. (2011). Haptic distal spatial perception mediated by strings: Haptic
“looming”, Journal of Experimental Psychology, 37(5), 1492-1511.
Cabe, P. A., & Pittenger, J. B. (1992). Time-to-topple: Haptic angular tau. Ecological
Psychology, 4, 241-246.
91
Carter, F. J., Frank, T. G., Davies, P. J., McLean, D. & Cuschieri, A. (2001).
Measurements and modeling of the compliance of human and porcine organs.
Medical Image Analysis, 5(4), 231-236.
Casper, J., & Murphy, R. R. (2003). Human-robot interactions during the robot-assisted
urban search and rescue response at the World Trade Center. IEEE Transactions
on Systems, Man, and Cybernetics-Part B: Cybernetics, 33, 367-385.
Chmarra, M. K., Dankelman, J., van den Dobbelsteen, J. J., & Jansen, F. W. (2008).
Force feedback and basic laparoscopic skills. Surgical Endoscopy and Other
Interventional Techniques, 22, 2140-2148.
Coles, T. R., Meglan, D., & John, N. W. (2011). The role of haptics in medical training
simulators: A survey of the state of the art. IEEE Transactions on Haptics, 4(1),
51-66.
Darby, M. L. (2000). Battlefield simulation: Building virtual environments. Journal of
Battlefield Technology, 3, 35–43.
Den Boer, K. T., Herder, J. L., Sjoerdsma, W., Meijer, D. W., Gouma, D. J., & Stassen,
H. G. (1999). Sensitivity of laparoscopic dissectors: What can you feel? Surgical
Endoscopy and Other Interventional Techniques, 13, 869-873.
Fung, Y.C. (1993). Biomechanics: Mechanical properties of living tissues. New York:
Springer.
Gomer, J. A., Dash, C. A., Moore, K. S., & Pagano, C. C. (2009). Using radial outflow to
provide depth information during teleoperation. Presence, 18(4), 304-320.
Gibson, E. (1953). Improvement in perceptual judgments as a function of controlled
practice of training. Psychological Bulletin, 50(6), 401-431.
Gibson, E. (1963). Perceptual Learning. Annual Review of Psychology, 14, 29-56.
Gibson, E. (1969). Principles of perceptual learning and development. New York:
Appleton-Century-Crofts.
Gibson, J. J. (1950). The perception of the visual world. Boston: Houghton Mifflin.
Gibson, J. J. & Gibson, E. (1955). Perceptual learning: Differentiation or enrichment?
Psychological Review, 62(1), 32-40.
Gibson, J. J. (1966). The senses considered as perceptual systems. Boston: Houghton
Mifflin.
92
Gibson, J. J. (1979). The ecological approach to visual perception. Boston: Houghton
Mifflin.
Hecht, H. & Savelsbergh, G. (2004) Advances in Psychology: Time-to-contact.
Amsterdam, The Netherlands: Elsevier Science.
Heck, R. H., Thomas, S. L., & Tabata, L. N. (2010). Multilevel and longitudinal
modeling with IBM SPSS. New York, NY: Routledge.
Heijnskijk, E. A. M., Pasdeloup, A., van der Pijl, A. J., Dankelman, J., & Gouma, D. J.
(2004). The influence of force feedback and visual feedback in grasping tissue
laparoscopically. Surgical Endoscopy and Other Interventional Techniques,
18(6), 980-985.
Hugh, T. B. (2002). New strategies to prevent laparoscopic bile duct injury – Surgeons
can learn from pilots. Surgery, 132, 826-835.
Klatzky, R. L., Lederman, S. J., & Matula, D. E. (1993). Haptic exploration in the
presence of vision. Journal of Experimental Psychology: Human Perception and
Performance, 19, 726-743.
Lamata, P., Gómez, E. J., Hernández, F., L., Pastor, A. O., Sánchez-Margallo, F. M., del
Pozo Guerrero, F. (2008). Understanding perceptual boundaries in laparoscopic
surgery, IEEE Transactions on Biomedical Engineering, 55(3), 866-873.
Long, L.O., Hartman, L., Pagano, C.C., Kil, I., Singapogu, R. & Burg (2014). Haptic
Perception of Distance-To-Break for Compliant Tissues in a Surgical Simulator.
Proceedings of the International Annual Meeting of the Human Factors and
Ergonomics Society, 1984-1988, Chicago, IL, Oct 27-31, 2014.
Long, L., Singapogu, J., DuBose, S., Arcese, G., Altenhoff, B., Burg, T., & Pagano, C.
(2012). A haptic simulator for training force skill in laparoscopic surgery.
Proceedings of the 2012 Interservice/Industry Training, Simulation, and
Education (IITSEC) Conference, Orlando, Florida.
Lee, D. N. (1976). A theory of visual control of braking based on information about timeto-collision. Perception, 5, 437-459.
Napieralski, P. E., Altenhoff, B. M., Bertrand, J. W., Long, L. O., Babu, S. V., Pagano, C.
C., Kern, J., & Davis, T. A. (2011). Near-field distance perception in real and
virtual environments using both verbal and action responses. ACM Transactions
on Applied Perception (TAP), 8(3), Article18.
93
Pagano, C. C. & Cabe, P. A. (2003). Constancy in dynamic touch: Length perceived by
dynamic touch is invariant over changes in media. Ecological Psychology, 15(1),
1-17.
Pagano, C. & Donahue, K. (1999). Perceiving the lengths of rods wielded in different
media. Perception & Psychophysics, 61(7), 1336-1344.
Perreault, J. & Cao, C. (2006). Effects of vision and friction and haptic perception.
Human Factors, 48(3), 574-586.
Peters, T. M., Linte, C. A., Moore, J., Bainbridge, D., Jones, D. L., & Guiraudon, G. M.
(2008). Towards a medical virtual reality environment for minimally invasive
cardiac surgery. Proc. Medical Imaging and Augmented Reality, 5128, 1-11.
Picod, G., Jambon, A. C., Vinatier, D., & Dubois, P. (2005). What can the operator
actually feel when performing a laparoscopy? Surgical Endoscopy and Other
Interventional Techniques, 19(1), 95-100.
Richardson, A. R., & Waller, D. (2005). The effect of feedback training on distance
estimation in virtual environments. Applied Cognitive Psychology, 19, 1089–
1108.
Rosen, J., Brown, J. D., De, S., Sinanan, M., & Hannaford, B. (2008). Biomechanical
properties of abdominal organs in vivo and postmortem under compression loads.
Journal of Biomechanical Engineering, 130.
Schiff, W., & Oldak, R. (1990). Accuracy of judging time to arrival: Effects of modality,
trajectory, and gender. Journal of Experimental Psychology: Human Perception
and Performance, 16, 303-316.
Shaw, B. K., McGowen, R. S., & Turvey, M. T. (1991). An acoustic variable specifying
time-to-contact. Ecological Psychology, 3(3), 253-261.
Singapogu, R. B., DuBose, S., Long, L. O., Smith, D. E., Burg, T. C., Pagano, C. C.,
Burg, K. J. L. (2013). Salient haptic skills trainer: Initial validation of a novel
simulator for training force-based laparoscopic surgery skills. Surgical
Endoscopy, 27(5), 1653-1661.
Singapogu, R. B., Pagano, C. C., Burg, T. C., Burg, K. J. L. (2011). Perceptual metrics:
Towards better methods for assessing for assessing realism in laparoscopic
simulators. Studies in Health Technologies and Informatics, 163, 588-590.
94
Singapogu, R.B., Pagano, C.C., Burg, T.C., Dorn, P.G., Zacharia, R., & Lee. D (2014).
Perceptually Salient Haptic Rendering for Enhancing Kinesthetic Perception in
Virtual Environments. Journal on Multimodal User Interfaces, 8, 319-331.
Singapogu, R. B., Smith, D. E., Altenhoff, B. M., Long, L. O., Prabhu, V. V., Pagano, C.
C., Burg, T.C., & Burg, K.L. (2012a). Assessing surgeon and novice force skill on
a haptic stiffness simulator for laparoscopic surgery. Studies in Health
Technologies and Informatics, 173, 469-474.
Singapogu, R. B., Smith, D. E., Long, L. O., Burg, T. C., Burg, K. J. L., Pagano, C. C.,
(2012b). Objective differentiation of force-based laparoscopioc skills using a
novel haptic simulator. Journal of Surgical Education, 69(6), 766-773.
Smyth, M. M., & Waller, A. (1998). Movement imagery in rock climbing: Patterns of
interference from visual, spatial, and kinesthetic secondary tasks. Applied
Cognitive Psychology, 12, 145-157.
Srinivasan, M. A., & LaMotte, R. H. (1995). Tactual discrimination of softness. Journal
of Neurophysiology, 73(1), 88-101.
Ström, P., Hedman, L., Särna, L., Kjellin, A., Wredmark, T., Felländer-Tsai, L. (2006).
Early exposure to haptic feedback enhances performance in surgical simulator
training: a prospective randomized crossover study in surgical residents. Surgical
Endoscopy and Other Interventional Techniques, 20, 1383-1388.
Tang, B., Hanna, G. B., & Cushieri, A. (2005). Analysis of errors enacted by surgical
trainees during skills training courses. Surgery, 240, 518-525.
Thompson, W., Willemsen, P., Gooch, A., Creem-Regher, S., Loomis, J., & Beall, A.
(2004). Does the quality of the computer graphics matter when judging distances
in visually immersive environments? Presence: Teleoperation and Virtual
Environments, 13(5), 560-571.
Tittle, J. S., Roesler, A., & Woods, D. D. (2002). The remote perception problem. Proc.
Human Factors and Ergonomics Society 46th Annual Meeting, Human Factors
and Ergonomics Society, 260-264.
Treverso, L. W. (1999). Risk factors for intraoperative injury during cholecystectomy:
An ounce of prevention is worth a pound of cure. Annals of Surgery, 229, 458459.
Van den Dobbelsteen, J. J., Schooleman, A., & Dankelman, J. (2007). Friction dynamics
of trocars, Surgical Endoscopy and Other Interventional Techniques, 21, 13381343.
95
Van der Meijden, O. A. J., & Schijven, M. P. (2009). The value of haptic feedback in
conventional and robot-assisted minimal invasive surgery and virtual reality
training: a current review. Surgical Endoscopy and Other Interventional
Techniques, 23(6), 1180-1190.
Vicentini, M., & Botturi, D. (2009). Human factors in haptic contact of pliable surfaces.
Presence, 18(6), 478-494.
Wagman, J.B., Shockley, K., Riley, M.A. & Turvey, M.T. (2001). Attunement,
calibration, and exploration in fast haptic perceptual learning. Journal of
Motivational Behavior, 33, 323-327.
Wagner, C. R., Stylopoulos, N., & Howe, R.D. (2002). The role of force feedback in
surgery: Analysis of blunt dissection. Proc, 10th Symposium on Haptic Interfaces
for Virtual Environment and Teleoperator Systems, 68-74.
Way, L. W., Stewart, L., Gantert, W., Kingsway, L., Lee, C. M., Whang, K., et al. (2003).
Causes and prevention of laparoscopic bile duct injuries – Analysis of 252 cases
from a human factors and cognitive psychology perspective. Annals of Surgery,
237, 460-469.
Westebring-Van Der Putten, E. P., Goossens, R. H. M., Jakimowicz, & J. J., Dankelman,
J. (2008). Haptics in minimally invasive surgery—a review. Minimally Invasive
Therapy, 17(1), 3-16.
Withagen, R. & Michaels, C. F. (2005). The role of feedback information for calibration
and attunement in perceiving length by dynamic touch. Journal of Experimental
Psychology: Human Perception and Performance, 31, 1379-1390.
Witmer, B. G., & Kline, P. B. (1998). Judging perceived and traversed distance in virtual
environments. Presence – Teleoperators and Virtual Environments, 7(2), 144167.
Yamada, H. (1970). Strength of biological materials. Baltimore, MD: Wiliams &
Wilkins.
Zhou, M., Perreault, J., & Schwaitzberg, S.D. (2008). Effects of experience on force
perception threshold in minimally invasive surgery. Surgical Endoscopy and
Other Interventional Techniques, 22, 510-515.
96
APPENDICES
97
Appendix A
Demographics Questionnaire
ID
Date
Age:
Sex: (circle one)
Male
Female
1. Do you currently have any problems with your hands, arms, or neck?
Yes
No
If yes, please describe:
2. Have you ever required surgery on your hands or arms (including fingers and wrists)?
Yes
No
If yes, please describe (including which hand or both):
3. Do you currently have any vision problems aside from corrected vision?
Yes
No
If yes, please describe:
4. Do you have any experience with videogames?
Yes
No
If yes, estimated past usage or current hours per week:
If yes, list/describe your 3 most commonly played games and their respective consoles.
5. Does this include first-person perspective games (e.g. first-person shooter)?
Yes
No
If yes, estimated past usage or current hours per week:
Please describe:
98
Appendix B
Effects of Friction
Table 14
T-values for Multiple Regression Analyses Predicting Break Point Estimates from Actual
Break Point Distance, Actual Break Point Force, and Friction Level
Participant
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
24
25
31
Actual
Distance
1.99
0.68
2.83**
-1.49
1.59
3.15**
1.33
12.07**
6.38**
-2.05*
1.78
1.83
4.01**
44.86**
10.51**
0.57
2.56*
3.71**
3.63**
39.29**
10.21**
7.21**
-0.9
1.98
12.67**
Pretest
Actual Force
Friction
-4.29**
-1.82
-4.87**
0.71
-2.23*
-3.56**
-2.75*
-1.16
-1.77
-2.74*
-0.81
-2.07*
-2.51*
-3.17**
-0.74
-2.97**
-2.85**
-0.74
-0.97
-2.1*
-1.57
0.02
0.33
-3.66**
-3.01**
-1.2
-3.44**
1.13
-1.73
0.94
-6.61**
-0.6
-1.94
-1.89
2.19*
0.003
0.38
-3.82**
-0.74
-4.24**
-2.74*
-2.34*
-1.78
-1.39
3.92**
1.42
-0.36
0.2
-3.13**
-1.05
*p < .05, **p < .01
99
Actual
Distance
1.53
6.99**
4.86**
4.28**
7.27**
8.78**
4.79**
5.85**
6.09**
13.52**
7.13**
5.04**
7.05**
74.68**
39.55**
4.68**
25.48**
8.15**
7.18**
91.89**
13.25**
58.15**
8.93**
8.53**
35.83**
Posttest
Actual
Force
0.28
-2.26*
-2.9**
-2.05
-2.53*
-2.19*
1.02
-1.73
-1.31
0.82
-0.03
-1.13
-1.66
-6.56**
-2.71*
-1.67
-2.8**
-1.26
-1.46
-4.49**
-0.38
-1.87
-2.39*
-2.32*
-1.94
Friction
-5.35**
-5.08**
-7.9**
1.84
-4.13**
-2.37*
-0.56
-3.83**
-1.92
-0.77
-2.16*
-1.74
-3.62**
-6.51**
-3.83**
-3.24**
-4.14**
-0.75
-2.22*
-4.21**
-3.46**
-2.53*
-1.79
-1.92
-1.23